Uploaded by lindondlovu

Microeconomics (competition, conflict and coordination) by Samual Bowels and Simon D Halliday

advertisement
2
MICROECONOMICS
Competition, Conflict, and Coordination
S A M U E L B O W L E S & S I M O N D. H A L L I D AY
S A M U E L B O W L E S & S I M O N D. H A L L I D AY
MICROECONOMICS:
COMPETITION,
C O N F L I C T,
& C O O R D I N AT I O N
2
This text is provisional and under active revision. Please do not quote without
permission.
Samuel Bowles (PhD, Economics, Harvard University) heads the Behavioral
Sciences Program at the Santa Fe Institute. He has taught microeconomic
theory to undergraduates and PhD candidates at Harvard University, the University of Massachusetts, and the University of Siena. He is the author or
co-author of Notes and Problems in Microeconomic Theory (1980), Microeconomics: Behavior, Institutions and Evolution (2005), and with the global
CORE team, The Economy (2017) and Economy, Society and Public Policy
(2019), both open-access introductions to economics (for majors and nonmajors respectively). His research has appeared in the American Economic
Review, Nature, Science, Journal of Political Economy, Quarterly Journal of
Economics, and Econometrica.
Simon Halliday (PhD, Economics, University of Siena, Italy) is an Associate
Professor in the Economics department at the University of Bristol. He has
Figure 1: Sam at the Santa Fe Institute
also taught graduate and undergraduate students at Smith College, the University of Cape Town and Royal Holloway, University of London. His research
in experimental economics, behavioral economics and economics education
has been published in the Journal of Economic Behavior and Organization,
Journal of Behavioral and Experimental Economics, the Journal of Economic
Education, and elsewhere.
Figure 2: Simon at Smith College
Contents
Preface
11
I
People, Economy and Society
17
1
Society: Coordination Problems & Economic Institutions
2
23
1.1
Societal coordination: The classical institutional challenge
. . . . . . . . . . . . . . . . . . . . . .24
1.2
The institutional challenge today . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27
1.3
Anatomy of a coordination problem: The tragedy of the commons . . . . . . . . . . . . . . . . . . .28
1.4
Institutions: Games and the rules of the game . . . . . . . . . . . . . . . . . . . . . . . . . . . . .30
1.5
Over-exploiting nature: Illustrating the basics of game theory . . . . . . . . . . . . . . . . . . . . .33
1.6
Predicting economic outcomes: The Nash equilibrium
1.7
Evaluating outcomes: Pareto-comparisons and Pareto-efficiency . . . . . . . . . . . . . . . . . . .41
1.8
Strengths and shortcomings of Pareto efficiency as an evaluation of outcomes . . . . . . . . . . . .43
1.9
Conflict and common interest in a Prisoners’ Dilemma . . . . . . . . . . . . . . . . . . . . . . . . .44
1.10
Coordination successes: An invisible hand game . . . . . . . . . . . . . . . . . . . . . . . . . . .49
1.11
Assurance Games: Win-win and lose-lose equilibria . . . . . . . . . . . . . . . . . . . . . . . . . .50
1.12
Disagreement Games: Conflict about how to coordinate . . . . . . . . . . . . . . . . . . . . . . . .53
1.13
Why history (sometimes) matters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .54
1.14
Application: Segregation as a Nash Equilibrium among people who prefer integration . . . . . . . .56
1.15
How institutions can address coordination problems . . . . . . . . . . . . . . . . . . . . . . . . . .61
1.16
Game theory and Nash equilibrium: Importance and caveats . . . . . . . . . . . . . . . . . . . . .63
1.17
Application: Cooperation and conflict in practice . . . . . . . . . . . . . . . . . . . . . . . . . . . .65
1.18
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .67
. . . . . . . . . . . . . . . . . . . . . . . .36
People: Self-interest and Social Preferences
2.1
Preferences, beliefs and constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .72
2.2
Taking risks: Payoffs and probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .78
2.3
Expected payoffs and the persistence of poverty . . . . . . . . . . . . . . . . . . . . . . . . . . . .81
2.4
Decision-making under uncertainty: Risk-dominance . . . . . . . . . . . . . . . . . . . . . . . . .84
2.5
Sequential games: When order of play matters . . . . . . . . . . . . . . . . . . . . . . . . . . . .87
2.6
First-mover advantage in a sequential game . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .89
2.7
Social preferences: Blame Economic man? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .91
2.8
Experiments on economic behavior
2.9
The Ultimatum Game: Reciprocity and retribution . . . . . . . . . . . . . . . . . . . . . . . . . . .95
2.10
A global view: Common patterns and cultural differences . . . . . . . . . . . . . . . . . . . . . . .98
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .93
71
4
3
4
5
2.11
The Public Goods Game . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
.
2.12
Application: Evidence from Public Goods Games . . . . . . . . . . . . . . . . . . . . . . . . . . 104
.
2.13
Social preferences are not "Irrational" . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
.
2.14
Application. The lab and the street . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
.
2.15
Application: A fine is a price . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
.
2.16
Complexity: diverse, versatile, and changeable people
2.17
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
.
. . . . . . . . . . . . . . . . . . . . . . . 110
.
Doing the best you can: Constrained optimization
3.1
Time: A scarce resource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
.
3.2
Utility functions and preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
.
3.3
Indifference curves: Graphing preferences
3.4
Marginal utility and the marginal rate of substitution . . . . . . . . . . . . . . . . . . . . . . . . . 126
.
3.5
Application: Homo economicus with Cobb-Douglas utility . . . . . . . . . . . . . . . . . . . . . . 132
.
3.6
The feasible set of actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
.
3.7
The marginal rate of transformation and opportunity cost . . . . . . . . . . . . . . . . . . . . . . 137
.
3.8
Constrained utility maximization: The mrs = mrt rule . . . . . . . . . . . . . . . . . . . . . . . . 140
.
3.9
The price-offer curve, willingness to pay, and demand . . . . . . . . . . . . . . . . . . . . . . . . 145
.
3.10
Social preferences and utility maximization
3.11
Application: Environmental trade-offs
3.12
Application: Optimal abatement of environmental damages . . . . . . . . . . . . . . . . . . . . . 154
.
3.13
Cardinal inter-personally comparable utility: Evaluating policies to reduce inequality . . . . . . . . 159
.
3.14
Application: Cardinal utility and subjective well-being . . . . . . . . . . . . . . . . . . . . . . . . 162
.
3.15
Preferences, beliefs, and constraints: An assessment . . . . . . . . . . . . . . . . . . . . . . . . 164
.
3.16
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
.
117
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
.
Property, Power, & Exchange: Mutual Gains & Conflicts
4.1
Mutual gains from trade: Conflict and coordination . . . . . . . . . . . . . . . . . . . . . . . . . . 172
.
4.2
Feasible allocations: The Edgeworth box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
.
4.3
The Pareto-efficient set of feasible allocations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
.
4.4
Adam Smith’s Impartial Spectator suggests a fair outcome . . . . . . . . . . . . . . . . . . . . . 182
.
4.5
Property rights and participation constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
.
4.6
Symmetrical exchange: Trading into the Pareto-improving lens . . . . . . . . . . . . . . . . . . . 190
.
4.7
Bargaining power: Take-it-or-leave-it . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
.
4.8
Application: Bargaining over wages and hours . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
.
4.9
Application. The rules of the game determine hours and wages . . . . . . . . . . . . . . . . . . . 200
.
4.10
First-mover advantage: Price-setting power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
.
4.11
Setting the price subject to an incentive compatibility constraint . . . . . . . . . . . . . . . . . . . 209
.
4.12
Application. Other-regarding preferences: Allocations among friends . . . . . . . . . . . . . . . . 212
.
4.13
The rules of the game and the problem of limited information . . . . . . . . . . . . . . . . . . . . 217
.
4.14
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
.
Coordination Failures & Institutional Responses
5.1
Common property resources, public goods, and club goods . . . . . . . . . . . . . . . . . . . . . 225
.
5.2
A common property resources problem: Preferences . . . . . . . . . . . . . . . . . . . . . . . . 228
.
171
223
5
5.3
Technology and environmental limits: The source of a coordination failure . . . . . . . . . . . . . 232
.
5.4
A best response: Another constrained optimization problem . . . . . . . . . . . . . . . . . . . . . 235
.
5.5
A best-response function: Interdependence recognized . . . . . . . . . . . . . . . . . . . . . . . 238
.
5.6
How will the game be played? A symmetric Nash equilibrium . . . . . . . . . . . . . . . . . . . . 241
.
5.7
How would the players get to the Nash equilibrium? A dynamic analysis . . . . . . . . . . . . . . 244
.
5.8
Evaluating outcomes: Participation constraints, Pareto improvements and Pareto-efficiency . . . . 247
.
5.9
A Pareto inefficient Nash equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
.
5.10
A benchmark socially-optimal allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
.
5.11
Government policies: Regulation and taxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
.
5.12
Private ownership: Permits and employment
5.13
Community: Repeated interactions and altruism . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
.
5.14
Application: Is inequality a problem or a solution? . . . . . . . . . . . . . . . . . . . . . . . . . . 274
.
5.15
Over-exploitation of a non-excludable resource . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
.
5.16
The rules of the game matter: Alternatives to over-exploitation . . . . . . . . . . . . . . . . . . . 282
.
5.17
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
.
II
Markets for Goods and Services
291
6
Production: Technology and Specialization
295
7
6.1
The division of labor, specialization and the market . . . . . . . . . . . . . . . . . . . . . . . . . 296
.
6.2
Production functions with a single input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
.
6.3
Economies of scale and the feasible production set . . . . . . . . . . . . . . . . . . . . . . . . . 300
.
6.4
Economies of scale, specialization and exchange . . . . . . . . . . . . . . . . . . . . . . . . . . 303
.
6.5
Comparative and absolute advantage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
.
6.6
Specialization according to comparative advantage . . . . . . . . . . . . . . . . . . . . . . . . . 311
.
6.7
History, specialization, and coordination failures . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
.
6.8
Application: The limits of specialization and comparative advantage
6.9
Production technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
.
6.10
Production functions with more than one input . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
.
6.11
Cost-minimizing technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
.
6.12
Technical change and innovation rents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
.
6.13
Application: What does the model of innovation miss? . . . . . . . . . . . . . . . . . . . . . . . . 334
.
6.14
Characterizing technologies and technical change . . . . . . . . . . . . . . . . . . . . . . . . . . 336
.
6.15
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
.
. . . . . . . . . . . . . . . . 317
.
Demand: Willingness to pay and prices
343
7.1
The budget set, indifference curves and the rules of the game. . . . . . . . . . . . . . . . . . . . 346
.
7.2
Income, prices and offer curves
7.3
Cobb-Douglas utility and demand
7.4
Application. Doing the best you can dividing your time . . . . . . . . . . . . . . . . . . . . . . . . 357
.
7.5
Application: Social comparisons, work hours and consumption as a social activity . . . . . . . . . 360
.
7.6
Quasi-linear utility and demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
.
7.7
Price changes: income and substitution effects . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
.
7.8
Application: Income and substitution effects of a carbon tax and citizen dividend . . . . . . . . . . 374
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
.
6
8
9
7.9
Application: Giffen Goods and The Law of Demand . . . . . . . . . . . . . . . . . . . . . . . . . 377
.
7.10
Market demand and price elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
.
7.11
Application. Empirical estimates of the effect of price on demand. . . . . . . . . . . . . . . . . . . 384
.
7.12
Consumer surplus and interpersonal comparisons of utility . . . . . . . . . . . . . . . . . . . . . 386
.
7.13
Application: The effect of a sugar tax on consumer surplus . . . . . . . . . . . . . . . . . . . . . 389
.
7.14
Application. Willingness to pay (for an integrated neighborhood)
7.15
Application: Market dynamics and segregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
.
7.16
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
.
. . . . . . . . . . . . . . . . . . 393
.
Supply: Firms’ costs, output and profit
405
8.1
Costs of production: An owner’s eye view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 408
.
8.2
Accounting profits and economic profits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
.
8.3
Cost functions: Decreasing and increasing average costs . . . . . . . . . . . . . . . . . . . . . . 411
.
8.4
Application: Evidence about cost functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
.
8.5
A monopolistic competitor selects an output level . . . . . . . . . . . . . . . . . . . . . . . . . . 417
.
8.6
Profit maximization: marginal revenues and marginal costs . . . . . . . . . . . . . . . . . . . . . 423
.
8.7
The markup, the price elasticity of demand, and entry barriers
8.8
Application: Evidence on the markup in drug prices . . . . . . . . . . . . . . . . . . . . . . . . . 433
.
8.9
Willingness to sell: Capacity constraints and market supply . . . . . . . . . . . . . . . . . . . . . 435
.
8.10
Economic profits and the market supply curve
8.11
Perfect competition among price-taking buyers and sellers: Shared gains from exchange . . . . . 440
.
8.12
The effects of a tax: Consumer surplus, profits, tax revenues and deadweight loss
8.13
Competition among price takers: An assessment . . . . . . . . . . . . . . . . . . . . . . . . . . 446
.
8.14
Two benchmark models of the profit-maximizing firm: Price takers and price makers.
8.15
Application: Dynamics – The growth of firms and the survival of competition . . . . . . . . . . . . 450
.
8.16
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
.
. . . . . . . . . . . . . . . . . . . 429
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . 438
.
. . . . . . . . 444
.
. . . . . . . 448
.
Competition, Rent-seeking & Market Equilibration
457
9.1
Modelling the continuum of competition: From one firm to many . . . . . . . . . . . . . . . . . . . 458
.
9.2
Reviewing the monopoly case, n = 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
.
9.3
Duopoly: Two firms’ best responses and the Nash equilibrium
9.4
Oligopoly and "unlimited competition": From a few firms to many firms . . . . . . . . . . . . . . . 472
.
9.5
The extent of competition and the markup over costs . . . . . . . . . . . . . . . . . . . . . . . . 476
.
9.6
Barriers to entry and the equilibrium number of firms . . . . . . . . . . . . . . . . . . . . . . . . 477
.
9.7
A conflict of interest: Profits, consumer surplus, and the degree of competition . . . . . . . . . . . 482
.
9.8
Limited competition and inefficiency: Deadweight loss
9.9
Coordination among firms: Duopoly and cartels . . . . . . . . . . . . . . . . . . . . . . . . . . . 487
.
9.10
Perfect price discrimination: Eliminating deadweight loss at a cost to consumers
9.11
Application: Price discrimination in action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494
.
9.12
Rent-seeking, price-making, and market equilibration . . . . . . . . . . . . . . . . . . . . . . . . 496
.
9.13
Application: When rent-seeking does not equilibrate a market – A housing bubble . . . . . . . . . 501
.
9.14
How competition works: The forces of supply and demand . . . . . . . . . . . . . . . . . . . . . 503
.
9.15
The "perfect competitor:" Rent-seeking firms competing in and for markets
9.16
Application: Declining competition and public policy . . . . . . . . . . . . . . . . . . . . . . . . . 510
.
. . . . . . . . . . . . . . . . . . . 463
.
. . . . . . . . . . . . . . . . . . . . . . . 483
.
. . . . . . . . . 490
.
. . . . . . . . . . . . 506
.
7
9.17
III
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
.
Markets with Incomplete Contracting
517
10 Information: Contracts, Norms & Power
523
10.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
.
10.2
Incomplete contracts: "... not everything is in the contract"
10.3
Principals and agents: Hidden actions and hidden attributes
10.4
Hidden attributes and adverse selection: The Lemons Problem . . . . . . . . . . . . . . . . . . . 530
.
10.5
Application: Health insurance
10.6
Hidden actions and moral hazards: A contingent renewal contract
10.7
The value of the transaction to the agent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
.
10.8
The agent’s best response: An incentive compatibility constraint . . . . . . . . . . . . . . . . . . 545
.
10.9
The principal’s cost minimization and the Nash equilibrium . . . . . . . . . . . . . . . . . . . . . 549
.
10.10
Short-side power in principal-agent relationships
10.11
A comparison with complete contracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 558
.
10.12
Features of equilibria with incomplete contracts: Summing up . . . . . . . . . . . . . . . . . . . . 562
.
10.13
Incomplete contracts and the distribution of gains from exchange
10.14
Application: Complete contracts in the gig economy . . . . . . . . . . . . . . . . . . . . . . . . . 568
.
10.15
Application: Norms in markets with incomplete contracts . . . . . . . . . . . . . . . . . . . . . . 570
.
10.16
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573
.
. . . . . . . . . . . . . . . . . . . . . 524
.
. . . . . . . . . . . . . . . . . . . . 527
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534
.
. . . . . . . . . . . . . . . . . 536
.
. . . . . . . . . . . . . . . . . . . . . . . . . . 555
.
. . . . . . . . . . . . . . . . . 564
.
11 Work, Wages & Unemployment
577
11.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577
.
11.2
Employment as a principal-agent relationship . . . . . . . . . . . . . . . . . . . . . . . . . . . . 578
.
11.3
Nash equilibrium wages, effort, and hiring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581
.
11.4
The employer’s profit-maximizing level of hiring . . . . . . . . . . . . . . . . . . . . . . . . . . . 584
.
11.5
Comparing the incomplete and complete contracts cases . . . . . . . . . . . . . . . . . . . . . . 590
.
11.6
Employment rents and the workers’ fallback option
11.7
Connecting micro to macroeconomics: A no-shirking condition . . . . . . . . . . . . . . . . . . . 599
.
11.8
Incomplete contracts & the distribution of gains from exchange . . . . . . . . . . . . . . . . . . . 603
.
11.9
Application: Contract enforcement technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . 604
.
11.10
Equilibrium unemployment and the wage curve . . . . . . . . . . . . . . . . . . . . . . . . . . . 607
.
11.11
The whole-economy model: Profits, wages, and employment . . . . . . . . . . . . . . . . . . . . 611
.
11.12
Monopsony, the cost of inputs and the level of hiring
. . . . . . . . . . . . . . . . . . . . . . . . . 595
.
. . . . . . . . . . . . . . . . . . . . . . . . 617
.
11.13
Monopsony and the cost of hiring (non-shirking) labor
11.14
The effects of a minimum wage on hiring and labor earnings . . . . . . . . . . . . . . . . . . . . 624
.
. . . . . . . . . . . . . . . . . . . . . . . 621
.
11.15
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 630
.
12 Interest, Credit & Wealth Constraints
635
12.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635
.
12.2
Evidence on credit and wealth constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 637
.
12.3
The wealthy owner-operator case
12.4
Complete credit contracts: A limiting case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
.
8
IV
12.5
The general case: incomplete credit contracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 650
.
12.6
The Nash equilibrium level of risk and interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654
.
12.7
Characteristics of the incomplete contract Nash equilibrium
12.8
Many lenders: Competition and barriers to entry
12.9
Wealth matters: Borrowing with equity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 667
.
12.10
Excluded and credit-constrained borrowers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 670
.
12.11
Why redistributing wealth may enhance efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . 672
.
12.12
Competition, barriers to entry and the distribution of rents . . . . . . . . . . . . . . . . . . . . . . 676
.
12.13
Application: From micro to macro: The multiplier and monetary policy
12.14
Application. Why cotton became king in the U.S. South following the end of slavery . . . . . . . . 684
.
12.15
Why and How Wealth Matters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 685
.
12.16
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 687
.
. . . . . . . . . . . . . . . . . . . . 659
.
. . . . . . . . . . . . . . . . . . . . . . . . . . 663
.
. . . . . . . . . . . . . . . 678
.
Economic systems and policy
691
13 A Risky & Unequal World
695
13.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 695
.
13.2
Choosing Risk: Gender differences
13.3
Risk preferences over lotteries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 699
.
13.4
Wealth differences and decreasing risk aversion . . . . . . . . . . . . . . . . . . . . . . . . . . . 703
.
13.5
Application: Risk, wealth and the choice of technology
13.6
Doing the best you can in a risky world . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 708
.
13.7
How risk aversion can perpetuate economic inequality
. . . . . . . . . . . . . . . . . . . . . . . 712
.
13.8
How insurance can mitigate risk and reduce inequality
. . . . . . . . . . . . . . . . . . . . . . . 714
.
13.9
Buying and selling risk: Two sides of an insurance market . . . . . . . . . . . . . . . . . . . . . . 720
.
13.10
Application: Free tuition with an income-contingent tax on graduates . . . . . . . . . . . . . . . . 725
.
13.11
Another form of insurance: A linear tax and lump sum transfer . . . . . . . . . . . . . . . . . . . 730
.
13.12
A citizen’s preferred level of tax and transfers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 735
.
13.13
Political rents: Conflicts of interest over taxes and transfers . . . . . . . . . . . . . . . . . . . . . 739
.
13.14
Application: Choosing justice, a question of ethics
13.15
Risk, uncertainty and loss aversion: Evaluation of the model . . . . . . . . . . . . . . . . . . . . 745
.
13.16
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 747
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 697
.
. . . . . . . . . . . . . . . . . . . . . . . 706
.
. . . . . . . . . . . . . . . . . . . . . . . . . 740
.
14 Perfect Competition & the Invisible Hand
751
14.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 751
.
14.2
A general competitive equilibrium
14.3
Market clearing and Pareto-efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 757
.
14.4
Prices as messages, markets as information processors
14.5
The Fundamental Theorems and Pareto efficiency
14.6
Perfectly competition and inequality: Distributional neutrality
14.7
Market failures due to uncompensated external effects . . . . . . . . . . . . . . . . . . . . . . . 772
.
14.8
Market dynamics: Getting to an equilibrium and staying there . . . . . . . . . . . . . . . . . . . . 775
.
14.9
Bargaining and rent-seeking: A more realistic model of market dynamics . . . . . . . . . . . . . . 778
.
14.10
Disequilibrium trading creates inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 781
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 753
.
. . . . . . . . . . . . . . . . . . . . . . 761
.
. . . . . . . . . . . . . . . . . . . . . . . . . 764
.
. . . . . . . . . . . . . . . . . . . . 766
.
9
14.11
Bargaining to an efficient outcome: The Coase Theorem . . . . . . . . . . . . . . . . . . . . . . 784
.
14.12
An example: How Coasean bargaining works . . . . . . . . . . . . . . . . . . . . . . . . . . . . 787
.
14.13
Application: Bargaining over a curfew . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 794
.
14.14
Bargaining, markets, and public policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 803
.
14.15
Application: Planning vs the market in the history of economics
14.16
Perfect competition or the perfect competitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 807
.
14.17
Conclusion: Ideal systems in an imperfect world . . . . . . . . . . . . . . . . . . . . . . . . . . . 808
.
. . . . . . . . . . . . . . . . . . 805
.
15 Capitalism: Innovation & Inequality
813
15.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 813
.
15.2
Capitalism’s success: The hockey stick of history . . . . . . . . . . . . . . . . . . . . . . . . . . 815
.
15.3
Capitalism and inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 817
.
15.4
Employment as insurance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 818
.
15.5
Explaining the hockey stick: Capitalist firms share risks & promote innovation
15.6
How can more equal societies also be innovative? . . . . . . . . . . . . . . . . . . . . . . . . . . 824
.
15.7
Measuring economic inequality: The Gini coefficient and the Lorenz curve . . . . . . . . . . . . . 826
.
15.8
Inequality and the macro-economy: A micro-economic explanation . . . . . . . . . . . . . . . . . 830
.
15.9
Market power and the distribution of income . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834
.
15.10
Modern monopoly, winners-take-all and public policy
15.11
Application: Public policy to raise wages and reduce unemployment and inequality . . . . . . . . . 839
.
15.12
Application: Trade unions, inequality, and economic performance . . . . . . . . . . . . . . . . . . 844
.
15.13
Capitalism as an economic and social order: Disparities in wealth and power . . . . . . . . . . . . 846
.
15.14
Would a wealth-poor person want to hold a risky asset? . . . . . . . . . . . . . . . . . . . . . . . 852
.
15.15
Risk, redistribution and innovation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 855
.
15.16
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 856
.
. . . . . . . . . . . 820
.
. . . . . . . . . . . . . . . . . . . . . . . . 836
.
16 Public policy and mechanism design
16.1
Mechanism design: Policy implementation by Nash equilibrium . . . . . . . . . . . . . . . . . . . 863
.
16.2
Optimal contracts: internalizing external effects of public goods . . . . . . . . . . . . . . . . . . . 866
.
16.3
The social multiplier of cigarette taxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 873
.
16.4
The theory of the second best and public policy . . . . . . . . . . . . . . . . . . . . . . . . . . . 879
.
16.5
Deception as an impediment to efficient exchange . . . . . . . . . . . . . . . . . . . . . . . . . . 882
.
16.6
When optimal contracts fail: The case of team production . . . . . . . . . . . . . . . . . . . . . . 887
.
16.7
The limits of incentives: Crowding out and crowding in. . . . . . . . . . . . . . . . . . . . . . . . 895
.
16.8
Beyond market versus government: Expanding the space for policies and institutions . . . . . . . 901
.
16.9
Application: A worker-owned cooperative. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 901
.
16.10
The distributional impact of public policies: Rent control . . . . . . . . . . . . . . . . . . . . . . . 904
.
16.11
Egalitarian redistribution to address market failures . . . . . . . . . . . . . . . . . . . . . . . . . 911
.
16.12
Why governments sometimes fail: A caveat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 912
.
16.13
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 913
.
861
Glossary
917
Glossary
919
10
17 End notes
937
Preface
To its 18th and early 19th century founders, the subject of economics was the
wealth of nations and people. This was no less true of Karl Marx, the most
famous critic of capitalism, than it was of Adam Smith’s whose The Wealth
of Nations is considered the most powerful defence of the then emerging
capitalist economic system.
Economics was at the time called political economy, and it sought to understand how and why society was being transformed as a result of capitalism,
a novel way of organizing how people produce, exchange and distribute the
things we live on. Capitalism continues to change the world, and the task of
economics is to understand this process, and how our economies might be
made to work better for people today and in the future.
Welcome to Microeconomics: Competition, Conflict, and Coordination and
best wishes for your journey through its content. Let’s begin by saying how we
came to think that economics is important and then explaining our strategy for
how you can best learn to do economics.
Economics engaged in the world
Contrary to its reputation among students for being remote from reality, economics has always been about changing the way the world works. The earliest economists – the Physiocrats in late 18th century France and the Mercantilists before them – were advisers to kings and queens of Europe. Today’s
macroeconomic managers, economic development advisors and advocates
of competing policies concerning intellectual property rights or the global
movements of goods and people continue this tradition of real world engagement. Economists have never been strangers to policy-making, constitution
building and attempts at economic reform for the betterment of people’s living
conditions.
Alfred Marshall’s (1842-1924) Principles of Economics, published in 1890 was
the first great text in what came to be called neoclassical economics. It opens
with these lines:
12
MICROECONOMICS
- DRAFT
“Now at last we are setting ourselves seriously to inquire whether .. there need
be large numbers of people doomed from their birth to hard work in order
to provide for others the requisites of a refined and cultured life, while they
themselves are prevented by their poverty and toil from having any share or
part in that life. ...[T]he answer depends in a great measure upon facts and
inferences, which are within the province of economics; and this is it which gives
to economic studies their chief and their highest interest.”
The hope that economics might assist in alleviating poverty and securing the
conditions under which free people might flourish is at once economics’ most
inspiring calling and its greatest challenge. Like many, both of us were drawn
to economics by this hope.
One of us (Simon) grew up in Cape Town under the system of racial segregation called apartheid. He vividly remembers the demonstrations that finally
brought that system down and the long lines of people waiting to vote in South
Africa’s first democratic elections in 1994. He volunteered in the poor townships surrounding Cape Town teaching critical thinking and debating, skills
required to make the new democracy work. . Having initially followed his passion for theater and poetry, he switched into economics to gain the analytical
tools to understand and address his country’s challenges.
The other of your authors (Sam), having been a schoolboy in India and a
secondary school teacher in Nigeria before turning to economics, naturally
came to the field expecting that it would address the enduring problem of
global poverty and inequality.
At age eleven Sam had noticed how very average he was among his classmates at the Delhi Public School – in sports, in school work, in just about
everything. A question that he then asked his mother has haunted him since:
"how does it come about that Indians are so much poorer than Americans,
given that as people we are so similar in our abilities?" And so he entered
graduate school hoping that economics might, for example, explain why workers in the United States produced almost as much in a month as those in
India produce in a year, and why the Indian population was correspondingly
poor.
We now know that the many conventional economic explanations for the gap
in standards of living between the two countries are part of the answer but
far from all of it: by any reasonable accounting, the difference in the amount
of machinery, land and other capital goods per worker and in the level of
schooling of the U.S. and Indian work forces explain much less than half of the
difference in output per hour of work.
It seems likely that much of the unexplained difference results from causes
that until recently have been less studied by economists but which are a
central theme of this book. Chief among these are differences in institutions,
that is differences in how the activities of the millions of actors in the two
C O N T E NTS
13
economies are coordinated by some combination of markets, private property,
social norms, and governments.
What should economics be about?
We do not think that we are atypical – either among our economics colleagues, or our students, or for that matter among people generally – in our
hope that economics can contribute to improving the way these institutions
work. The CORE Team – a group of economic researchers and teachers who
have created an open access introductory economics course (www.coreecon.org) – posed the following question to students around the world on
the first day of their introductory classes: "what is the most pressing problem
economists today should be addressing?" The results from a total of 4,442
students from 25 universities in twelve countries over the years 2016-18 are
summarized in the word cloud in Figure 3.
The themes are remarkably consistent across universities and countries.
Unemployment, inflation, and growth, all important topics in most macroeconomics courses, are on the minds of students. But inequality (along with
"poverty") is the overwhelmingly dominant issue. Environmental sustainability
(and "climate change"), the future of work (robots, digitalization), globalization
and migration, innovation, financial instability; and how governments work
("corruption," "war") are also present. A few students also identified particular
political events, like the "Brexit" referendum in 2016, that favored the U.K. leaving the European Union, as problems that economics should address. In word
clouds based on more recent surveys "climate change" is as large a concern
as "inequality" among students.
Figure 3: Student replies to the question "What
is the most pressing problem economists
should be addressing?" The size of the font
is proportional to the frequency with which
subjects mentioned the word or term. Surprisingly
professional economists at the New Zealand
Treasury and central bank and new hires at
the Bank of England responded very similarly
to students.The less frequently mentioned –
smaller font– topics are more readable in the
individual word clouds from each of the 25
samples of students that you can access at
https://tinyco.re/6235473
The microeconomic theory that you will learn has a lot to say about these
issues. Included are tried and true workhorse concepts that you have probably
14
MICROECONOMICS
- DRAFT
already encountered, like opportunity costs, mutual gains from exchange,
constrained optimization and trade offs. Also essential in understanding
issues like those in the word cloud are concepts that have more recently
risen to prominence among economists. Examples are the importance of
cooperation and social (rather than entirely selfish) motivations and modeling
strategic interactions among people, including conflicts over the distribution of
the mutual gains from exchange.
"If you are not doing something, you are not learning anything!"
The phrase just above is our motto when it comes to learning. Economics
is not just something you learn. It is something you do. Think of studying
economics as learning a new language. Mastering a large vocabulary and
the grammatical rules is essential, but it is not the same as speaking the
language.
The test of what you have learned after studying this book is not just what you
know, but what you can do with it. Doing economics is what you can say or
write – the case you can make for or against a proposed economic policy, the
analysis of the reasons for some new development in the global economy – in
other words what you can do as a result of what you know.
Like mastering a new language, doing economics is essential to learning the
subject. And also like a language, you will learn to do economics more readily
if you have a clear need to know.
We begin each chapter with a real world problem or example that can be
better understood using the concepts and models to be introduced in the
chapter. These opening paragraphs suggest the need to know what is to
follow. The empirical examples also serve as a reminder that the point to the
model is to understand the world; and as we proceed through chapters we will
ask: how good a job does this particular model do in that respect?
In the margin at the beginning of each chapter is a set of learning objectives
phrased as new capacities to do things that most likely you were unable
to do before. We place great emphasis on your ability to solve problems in
which there are right and wrong answers. But it is also important to learn
how to formulate arguments and hypotheses about questions that are thus
far unanswered, some of which may remain so, and to express economically
informed opinions on issues that will continue to be debated due to the fact
that people’s values differ.
Interspersed with the contents of the chapters, but offset by boxes, are two
important resources:
Mathematics Notes M-notes contain the details of mathematical derivations
C O N T E NTS
and other analysis as well as worked examples that illustrate the mathematical models in the text.
Checkpoints are self-tests to confirm that you understand the content of the
section. The first step in "doing economics" is by checking your understanding of the passage you have just read.
At the end of each chapter you will find the following:
Important Ideas The main ideas in each chapter are provided in a table. At
the end of the book, you will also find that all the definitions of the book are
included as a glossary for you to consult and improve your understanding.
Mastering the use of these terms is essential to doing economics. Try
using each of them in a complete sentence of our own.
Making connections provides some guidance in seeing how the ideas in
each chapter are connected to each other and to other themes in the book,
so that you will be able to draw together the ‘big picture’ about the main
messages and themes of the book. Try restating these connections making
use of the terms in Important Ideas. Or better yet: make a mind map using
the Important Ideas and Making Connections features.
Mathematical Notation The book contains a variety of important mathematical
tools to help model the various economic ideas in the book. To assist you
with your reading of each chapter and to understand better each model
you encounter, we provide a table of the mathematical notation you will
encounter in that chapter. There is also a complete list of the notation used
at the end of the book.
We use the margins of the book for a variety of purposes:
Definitions We define important terms in the margins where they first are
introduced. All of the definitions are collected in a glossary.
Reminders We put reminders in the text often to help you to see the connections of ideas throughout the book.
Example An example will often illustrate an idea with a relevant example
of a person, firm, or country making decisions that are similar to those
described in the text.
Fact Check When we need to verify or illustrate an idea with data or an
empirical example we will do so with a Fact Check.
History These introduce you to some of those people who have contributed to
economics or to relevant historical facts.
M-check If an idea requires a brief mathematical clarification that does not
require it’s own M-Note, then we may convey that in a margin note.
15
16
MICROECONOMICS
- DRAFT
Economics is an integrated body of knowledge, and it is best learned in a cumulative way, mastering a set of concepts and going on to use those concepts
in mastering additional concepts. What this means, practically is that it is best
to study earlier chapters before moving on to later ones. Sections labeled
"application" however provide illustrations of how the ideas and models being
taught in a particular chapter can be used, and these do not introduce new
material that is essential to the chapters that follow.
Micoeconomics is waiting for you. Just do it!
Samuel Bowles and Simon Halliday
Santa Fe Institute, Santa Fe, New Mexico, U.S.A, and Smith College, Northampton, Massachusetts, U.S.A.
Part I
People, Economy and Society
19
The man . . . enamored of his own ideal plan of government, . . . seems to
imagine that he can arrange the different members of a great society with as
much ease as the hand arranges the different pieces upon a chess-board . . . but
. . . in the great chess-board of human society, every single piece has a principle
of motion of its own, altogether different from that which the legislature might
choose to impress upon it.
If those two principles coincide and act in the same direction, the game of
human society will go on easily and harmoniously, and is very likely to be happy
and successful. If they are opposite or different, the game will go on miserably,
and the society must be at all times in the highest degree of disorder.
Adam Smith,Theory of Moral Sentiments, 1759,Part VI, Section 1
As individuals, our physical capacities are hardly remarkable compared to
other animals. But by coordinating with others – finding ways that our individual efforts can add up to a whole that is more than the sum of its parts –
humans are unique as a species, engaging in common pursuits on a global
scale and for better or worse, transforming nature and inventing previously
unimagined devices and ways of life. Economics provides a lens for studying
this social aspect of human uniqueness by analysing how people interact with
H I S TO RY What makes humans unique
among all the animals is our capacity to cooperate in very large numbers
and to adjust the ways that we cooperate to changing circumstances. Here
(https://tinyurl.com/y3bpy4px) the Israeli
historian Yuval Noah Harari explains why this
is so.
each other and with our natural surroundings in producing and acquiring our
livelihoods.
We begin (in Chapter 1) by developing a common framework for studying
the various types of social interactions using game theory to pose a question older than economics. This is: how can a society’s institutions – its laws,
unwritten rules and social norms – harness individuals’ pursuit of their own
objectives to generate common benefits and to avoid outcomes that none
would have chosen. The challenge is how to combine freedom – individuals’ pursuit of their own objectives – with the common good, improving the
livelihoods of all members of the society.
This challenge is called the problem of societal coordination: how we can coordinate – that is organize – our actions to yield desirable results for society?
The example of societal coordination we use in Chapter 1 to illustrate this
challenge is about the other aspect of economics: how we relate to our natural
surroundings, illustrated by a problem of over-exploiting an environmental
resource.
Adam Smith, considered by many to be the founder of economics, understood
the challenge well. And he understood that economics – or "political economy" as it was then called –is fundamentally a social science: it is about how
people interact. You can see this in his warning above about the disastrous
consequences of treating people as if they were simply chess pieces who
could be moved around on the chess board of life at the will of a government,
more or less like an engineer might design a machine.
An adequate response to the challenge of combining freedom and the com-
E CONOMICS is the study of how people
interact with each other and with our natural
surroundings in producing and acquiring our
livelihoods.
20
MICROECONOMICS
- DRAFT
mon good must therefore be based on knowledg of how individuals act depending on the situation they are in, and how changing the situation will
change how they act. We therefore (Chapter 2) turn to individuals and their
motives– whether self interested or generous, opportunistic or ethical – explaining how people do the best they can in given situations. In this chapter
we consider individuals in situations where they act in isolation rather than
interacting with other individuals.
But people rarely act in isolation: Economics allows us to understand the
sometimes surprising or unintended society-wide effects of when we interact
with others, whether it be directly with our own employer or indirectly with
literally millions of people involved in producing and distributing the goods
making up our livelihoods.
A basic insight for this understanding is that we are better off by interacting
with others. But our interactions also give rise to conflicts. When people
engage with others in buying and selling, working and investing there are
mutual benefits potentially available to all parties involved. This must be the
case if participation in these and other economic activities is voluntary. But
unavoidably there are also conflicts over how these mutual gains are divided
(Chapter 4).
We evaluate the outcomes of economic interactions by two standards:
• Efficiency : the extent to which all of the potential gains are realized (which
is how economists use the term efficiency) and
• Fairness: whether the distribution of the gains and the process that determines who gets what is just.
And we study the various ways that exchanges and other economic activities
may be carried out and how they may affect both the efficiency and fairness of
the outcome.
In our interactions with each other and with nature we frequently fail to exploit
all of the potential mutual gains. An example is when a person with the capacity and desire to produce goods and services needed by others cannot find
a job. Another is over-exploitation of a fishery or some other environmental
resource. These are called coordination failures because they result from
inadequacies in the ways that our institutions coordinate the ways that we
interact.
Coordination failures are often due to our conflicts over the distribution of potential mutual gains or to the fact that when we act we do not take account of
the effects of our actions on others (Chapter 5). Markets, government policies,
well-designed property rights, a concern for one another’s well being, and
communities can help address these coordination failures so that no potential
mutual gains remain unexploited, and the distribution of gains is regarded as
21
fair. The final chapter of this book will bring together the concepts, models
and other ways of thinking that you have learned and apply them to the challenge of improving the way the economy works by both of these standards:
efficiency and fairness.
1
Society: Coordination Problems & Economic Institutions
DOING ECONOMICS
Two neighbors may agree to drain a meadow, which they possess in common;
because ’tis easy for them to know each others mind; and each must perceive,
that the immediate consequence of his failing in his part, is the abandoning of
the whole project.
But ’tis very difficult and indeed impossible, that a thousand persons shou’d
agree in any such action; it being difficult for them to concert so complicated a
design, and still more difficult for them to execute it; while each seeks a pretext
to free himself of the trouble and expense, and wou’d lay the whole burden on
others.
David Hume, A Treatise of Human Nature, Volume II (1967 [1742]: 304)
At the turn of the present century, the process of economic development had
bypassed almost all of the two hundred or so families that made up the village
of Palanpur in the Indian state of Uttar Pradesh. But for the occasional watch,
bicycle or irrigation pump, Palanpur appeared to be a timeless backwater,
untouched by India’s cutting edge software industry and booming agricultural
regions. Less than a third of the adults were literate, and most had endured
the loss of a child to malnutrition or to illnesses that had long been forgotten in
other parts of the world.
A visitor to the village approached a farmer and his three daughters weeding
a small plot of land. The conversation turned to the fact that Palanpur farmers
This chapter will enable you to::
• Use game theory to analyze how people
interact in the economy, each affecting
the conditions under which the others
decide how to act.
• Understand why the outcomes of
interactions are often worse for people
than they could be and how interactions
might be better organized to improve the
quality of people’s lives.
• Recognize that unsatisfactory outcomes
occur when people fail to coordinate with
each other and to take account of the
effect of their own actions on others.
• Explain how problems like environmental
damage and global poverty can be the
result of failed coordination.
• Represent institutions as "the rules of the
game."
• Identify that economic institutions
determine incentives for people’s
behavior and can affect how successfully
we address coordination problems.
• Explain why when people have limited
information and conflicts of interest
they often fail to implement ’win-win’
outcomes.
plant their winter crops several weeks after the date that would maximize the
amount of grain they could get at harvest time. The farmers knew that planting
earlier would produce larger harvests, but no one, the farmer explained, wants
to be the first farmer to plant their seeds, as the seeds on any lone plot would
be quickly eaten by birds.
Curious, the visitor asked if a large group of farmers, perhaps members of the
Figure 1.1: Palanpur farmers threshing and
winnowing grain (separating grain from chaff) .
Photo courtesy of Nicholas Stern.
24
MICROECONOMICS
- DRAFT
same extended family, had ever agreed to plant their seeds earlier, perhaps
on the same day to minimize the individual losses. “If we knew how to do
that,” the farmer said, looking up from his hoe and making eye contact with the
visitor for the first time, “we would not be poor.”1
1.1 Societal coordination: The classical institutional challenge
For the Palanpur farmers, the decision when to plant is a coordination problem. A coordination problem is a situation in which people could all be better
off, or at least some be better of and none be worse off, if they all jointly decided how to act – that is, if they coordinated their actions – than if they act
individually.
The planting choice is a coordination problem because:
• the farmer does better or worse depending on what other farmers do,
• all the farmers would do better if they could coordinate their actions by
jointly agreeing to all do what would be mutually beneficial namely, planting
early, but
• it is a problem because the farmers may not be able to coordinate, and
• if they do not coordinate, then all of the farmers will do worse than they all
could otherwise have done (that is, had they all planted late).
To stress the fact that coordination problems often affect an entire population
(even though we explain them using two person examples) we sometimes
use the expression societal coordination problems. Notice that one farmer
cannot dictate the actions of the other farmers, nor can they come to a common agreement about what to do ("if we knew how to do that, we would not
be poor") – the inability to come together and coordinate is at the heart of
coordination problems.
Our example is about farming, but it could just as well have been about the
owners of many firms each producing some different product deciding independently whether to invest in new buildings and equipment. Firms will only
invest only if they anticipate that there will be sufficient demand for the resulting increase in their outputs. Each firm’s investment – purchasing new
machinery and construction materials, and hiring more employees – means
greater demand for the other firms’ products, including the products purchased by the newly employed workers constructing the new offices, factories
and the like with their earnings.
If they all invest, then the other firms’ investments will create sufficient demand
to purchase the output from each of the firms’ expanded capacity. But if one
firm expands and the rest do not, then that firm is likely to find that it cannot
sell all of what it is now capable of producing.
C OORDINATION PROBLEM A coordination
problem is a situation in which people could
all be better off (or at least some be better of
and none be worse off) if they jointly decide
how to act – that is, if they coordinate their
actions – than if they act independently. For
example, deciding on which side of the road
to drive is a coordination problem.
H I S TO RY In his address accepting the Nobel
Prize for economics in 1979, University of
Chicago economist T.W. Schultz said: "Most
of the people in the world are poor, so if we
knew the economics of being poor, we would
know much of the economics that really
matters." He was right then and he is right
now.2
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
The best choice for each firm depends on the choices made by other firms. If
they could coordinate – decide jointly that they would all invest – they would
all profit, but coming to such an agreement may not be possible. The owners
of firms therefore face a coordination problem similar to the problem faced
by the Palanpur farmers. Their individual success depends on their ability to
coordinate their actions with others. This is not a new problem: it was on a
central concern of the founders of economics.
David Hume (the 18th century British philosopher and economist quoted
at the start of this chapter) used an example – two landowners considering
draining a meadow – to pose what he considered the most important problem
facing society, namely, devising institutions that would reconcile the pursuit
of individual objectives (avoiding the "trouble and expense" in his example
of the meadow) with getting desired societal outcomes (improving the value
of the meadow by draining it). His simple two-person example was meant to
illustrate the need (in a society of "a thousand persons") for a government to
address the broader societal coordination problems of his day.
Though the term was invented only two centuries after Hume, he was using
what we now call game theory to make his case. Let’s apply his reasoning to
the farmers of Palanpur. Like Hume we will consider just two farmers as a way
of representing the institutional challenge faced by the entire village.
Figure 1.2 shows the outcomes for two players, Aram and Bina, choosing
when to plant their grain. The figure illustrates the values of the farmers’
crops, which, in a poor village like Palanpur, is the main incentive for farmers,
whether they consume the crop themselves or sell it for money income to
spend on other things. Each farmer can either plant early or plant late and
while (also as in Hume’s example) two people could probably come to some
agreement about what to do, remember that we are using this two-person
example to illustrate the entire village of about 200 families of farmers, so we
assume that they cannot coordinate on some agreed upon actions for the two
jointly. There are four possible outcomes:
• If both players plant early, they each achieve their best possible harvest,
because they grow the most grain through sharing the risk of having their
seeds eaten by birds (outcome c in Figure 1.2).
• If Aram plants early while Bina plants late, Aram has his seeds eaten by
birds and gets no harvest (the worst outcome for him), whereas the late
planter gets a good (but not the best) harvest. While none of her seeds are
eaten by the birds, planting late is not the best for growing the most grain
(outcomes b and d in Figure 1.2). The same is true if Bina planted late
when Aram planted early.
• If both plant late, they harvest a smaller crop while also sharing the risk of
their seeds being eaten, a bad outcome (outcome a in Figure 1.2).
& ECONOMIC INSTITUTIONS
25
26
MICROECONOMICS
- DRAFT
Figure 1.2: Planting in Palanpur. This figure
shows "what-if" outcomes for planting in Palanpur.
Each column represents a possible combination of
Aram planting early or late and Bina planting early
or late with the corresponding outcomes being
worst, bad, good, or best in terms of how much
grain they grow.
The people of Palanpur are stuck in the bad outcome even though they would
all be better off if they all planted early (they would both move from a "bad"
outcome to the "best" outcome in the figure). They are experiencing a co-
C OORDINATION FAILURE A coordination
failure occurs when people facing a coordination problem fail to coordinate their actions
in a way to implement outcome that allows
them all to be better off (or at least some to
be better off and none to be worse off).
ordination failure, namely a coordination problem that is not addressed by
appropriate institutions. A modern day David Hume would point out that a government could simply impose a sufficient tax on those planting late to ensure
that most farmers would plant early.
Adam Smith, a generation after Hume, would stress the value of the exchange
of privately owned goods on competitive markets as a way of coordinating
the actions of large numbers of people, who would be guided (even without
knowing it) by what he termed "an invisible hand." Hume, Smith and the other
founders of European political philosophy and political economy posed what
we call the classical institutional challenge.
These philosophers and economists wanted to know how to design institutions – rules and practices governing peoples’ behavior – so that people
could be left free to make their own decisions, and at the same time avoid
outcomes that were inferior for everyone. More precisely, how do we design
institutions which encourage coordination by free choice while avoiding poor
outcomes such as planting late in Palanpur? The 18th and 19th century political economists and philosophers who founded the field of economics were
H I S TO RY Adam Smith wrote the following:
“[E]very individual [. . .], indeed, neither
intends to promote the public interest, nor
knows how much he is promoting it [. . .] he
intends only his own security; . . . he intends
only his own gain, and he is in this . . . led by
an invisible hand to promote an end which
was no part of his intention . . . By pursuing
his own interest he frequently promotes that
of the society more effectually than when he
really intends to promote it.”3
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
& ECONOMIC INSTITUTIONS
attempting to provide solutions to coordination problems.
Checkpoint 1.1: Planting in Palanpur: A Coordination Problem
Imagine that you are Bina in the figure above, and that you did not know
whether Aram would plant early or late. What would you do? Suppose you
and Aram were neighbors and you could talk with him; what would you say?
1.2 The institutional challenge today
The classical institutional challenge remains with us, although some of the
forms that it takes today – including global climate change and the appropriate
intellectual property rights for sharing digitized knowledge – were unknown to
the great 18th and 19th century thinkers.
Consider the following coordination problems:
• How do we sustain the global environment? To avoid damaging climate
change we need to coordinate our reduction of emissions. Many people
and firms would prefer that someone else reduce their carbon footprint.
How can we address climate change in a way that is both fair and imposes
the least possible costs?
• How do we make the best use of our ability to create and use knowledge?
If we all agree to share the knowledge we have with others we may all be
better off: when I transfer my knowledge to you I do not lose the ability to
continue using it. But each of us may profit by restricting others’ use of our
knowledge by means of patents, copyrights and other intellectual property
rights.
• How do we move around a city without overcrowding streets and causing
delays? My decision whether to drive, walk, or take public transport affects
not only my own travel time, but also the degree of traffic congestion and
delays experienced by everyone else. Everyone might be better off if the
use of private vehicles was substantially reduced, but few will reduce their
driving unless other people reduce theirs as well.
These are all coordination problems because an outcome that is better for all
is possible if people find a way to jointly agree to a course of action. But for
reasons we will explain in detail, people also routinely fail to coordinate and
suffer bad consequences as a result, including the following:
• over use of some resources illustrated by pollution, over-grazing, traffic
congestion, and climate change; and
• under use of other resources such as the productive capacities and creativity of people and the knowledge that we have created, illustrated by
Figure 1.3: Traffic headed out of a major city.
Image Credit: Photo by Preillumination SeTh
(@7seth).
27
28
MICROECONOMICS
- DRAFT
unemployment and the enduring poverty of the people of Palanpur and
villages like it around the world.
Checkpoint 1.2: Coordination Problems You Have Known
Think of a social interaction in which you have been involved that was a coordination problem and using the description of why planting in Palanpur is a
coordination problem (the bulleted points above) explain why it was a problem
and how coordination might have (or did) address the problem.
1.3 Anatomy of a coordination problem: The tragedy of the commons
The over use of environmental resources provides a good illustration of why
coordination problems arise.
In 1968, Garrett Hardin, an ecologist, famously described what he called
the tragedy of the commons, an example of a coordination failure.4 He
told a story about a group of herders who share a pasture. The pasture was
common land – hence a “commons” – shared by many herders. But why was
his story a tragedy?
Each herder could put as many animals in the pasture as they wished, and
overgrazing will lead to erosion and the ruin of the pasture. Hardin reasoned
that if the land is common to all and no one herder owns it, each herder has
no interest in limiting how many animals they put in the common pasture. A
ruined pasture is of no value to any of the herders. But each herder’s selfinterest leads them to neglect the effect their actions have on others. The
outcome is a tragedy.
With the term tragedy of the commons, Hardin gave social science one of the
most evocative metaphors since Adam Smith’s “invisible hand.” Indeed Hardin
called his tragedy a “rebuttal to the invisible hand.” The two metaphors are
powerful because they capture two essential yet contrasting social insights.
When guided by an invisible hand, social interactions reconcile individual
choice and socially desirable outcomes. By contrast, the actors in the tragedy
of the commons pursue their private objectives to tragic consequences for
themselves and others.
The natural setting for Hardin’s tragedy was chosen for its imagery. The underlying problem applies to many situations where people typically cannot or
do not take account of the effects of their actions on the well-being of others.
You can think of a city’s streets as a commons, and people deciding to drive
rather than walk, bike, or use public transport as similar to the herders putting
cattle on the common. The modern day "tragedy of the roadways" is a traffic
jam.
T RAGEDY OF THE C OMMONS The tragedy
of the commons is a term used to describe
a coordination failure arising when a shared
resource available for all to use (’the commons’) is over-used so that all users are
worse off than they would have been if they
had coordinated their actions so that use
was restricted.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
& ECONOMIC INSTITUTIONS
29
What are the common elements in Hume’s drain the meadow problem, the
farmers in Palanpur planting late, Hardin’s herders overgrazing their pasture
and our modern city dwellers clogging the streets with their vehicles?
In each of these three cases, the reason why uncoordinated activities of
people pursuing their own ends produce outcomes that are worse for all is
that each participant’s actions affect the well-being of others but these effects
are not taken into account by the individual actors when they decide how to
act. These impacts of our actions on others that we do not take account of in
deciding what to do are termed external effects.
Here are the external effects that actors in our four examples do not take into
E XTERNAL E FFECT An external effect occurs
when a participant’s action confers a benefit
or imposes a cost on other participants and
this cost or benefit is not taken into account
by the individual taking the action. External
effects are also called simply externalities.
External effects that result in costs to others
are called negative external effects or
external diseconomies. External effects that
confer benefits on others are called positive
external effects, external benefits, or external
economies.
account when deciding what to do:
• The person who lives in a city who drives to work, adds congestion to the
streets, and therefore increases the travel time of others.
• Hume’s farmer who does not drain the swamp and imposes the cost of
doing so on the other farmer.
• The Palanpur farmer who plants late, imposes a cost on the other farmer
who will have his seeds devoured by birds if he plants early. Likewise the
farmer who plants early confers a benefit on the other farmer who can
benefit by planting at the right time (early) without severe losses of seed to
the birds.
• The herder who places additional cattle on the common pasture reduces
the grass available to the other herders stock.
Addressing coordination problems by internalizing external effects
Simply abolishing these and other external effects that are the root of coordination problems is not an option. There is no way to organize society so
that nothing that we do would affect others, each person on his or her self
sufficient island.
Apart from not being much fun, life would be impossible in a society of total
social isolates (just think about how the next generation would be born and
raised!). So, to address the classical institutional challenge as to prevent
or at least minimize coordination failures we need to find ways of inducing
each participant to take adequate account of the effects of their actions on
others.
This is called internalizing an external effect. We use the term external effect
because the effect is outside of the individual’s process of decision-making
when taking the action. To internalize the external effect, you ensure that
the person who acts bears the costs of their negative effects on others and
reaps the rewards of their positive effects on others. In this way the otherwise
I NTERNALIZATION OF EXTERNAL EFFECTS
in economics refers to any way that people
can be brought to take appropriate account
of the effects of their actions on others. In
psychology the term internalization means
to to adopt societies values or standards as
one’s own values.
30
MICROECONOMICS
- DRAFT
"external" costs and benefits become part of the individual’s decision-making
process, leading them to "take adequate account of the effects of her actions
on others."
If the “others” are our family, our neighbors, or our friends, our concern for
their well-being or our desire to be well regarded by others might get us to
take account of the effects of our actions on them. Reflecting this fact, an
important response to the classical institutional challenge – one that long
predates the classical economists – is that caring for the well-being of others
need not be confined to friends and relatives but may extend to all of those
with whom we interact. Ethical guides such as the “golden rule” are ways that
H I S TO RY The “golden rule” is “to do unto
others as you would have them do unto you"
(Matthew, 7: 12). Or, treat others as you
would like to be treated yourself. The same
ethical principle is found in Islamic scriptures
and in the teaching of other religions.
people often internalize the effects of our actions on others, even when the
others are total strangers to us.
But, over the past five centuries, people have come to interact not with a few
dozen people as humans have for most of our history and pre-history but
directly with hundreds and indirectly with millions of strangers. The classical
economists in the 18th century were responding to the fact that the generosity
or ethical motivations that one might feel towards ones family or neighbors
would not be sufficient to induce people to take account of the effect of their
actions on others once these external effects spread across the entire network
of global interactions.
An objective that economics has set for itself from that day until today, therefore, has been to design and implement institutions that would induce people
to act as if they cared about those who were affected by their actions even
when that was not literally true.
Checkpoint 1.3: External effects
a. Provide an example of a negative external effect that occurs in a social
interaction. Explain why it is negative and why it is external.
b. Provide an example of a positive external effect that occurs in a social interaction. Explain why it is a positive external effect.
1.4 Institutions: Games and the rules of the game
Institutions
Institutions are the laws, norms, and beliefs that influence how people interact,
and what the outcomes of these interactions will be.
People adopt the be-
haviors prescribed by institutions (e.g. drive on the right if you are in the U.S.)
because of some combination of
• laws enforced by a government (you will be arrested and fined for driving
on the left in Brazil, the U.S, France, and other countries where driving on
the right is the law.)
I NSTITUTIONS Institutions are the laws,
informal rules, and conventions which
regulate social interactions among people
and between people and the biosphere.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
& ECONOMIC INSTITUTIONS
31
• social pressures – sometimes termed informal rules because they are not
enforced by governments (your friends and neighbors will disapprove and
think less of you if you drive on the left), and
• information that you have about what others will do (you expect others to
drive on the right, so you will avoid accidents by doing the same.)
The late Nobel Laureate Douglass North called institutions the “rules of the
game.”5 People can change these rules, so institutions can themselves be outcomes of games that govern how the rules of the game can be changed.
Because institutions are the rules of the game for how people (and businesses, and trade unions, and governments) interact, we now introduce Game
theory. Game theory uses mathematical models and verbal arguments to
analyze how the outcomes of the interaction for the participants will depend
on the rules of the game and the objectives of the players. It has been used
extensively in economics and the other social sciences, biology, and computer
science.
Game theory focuses on strategic interactions where participants are interdependent and are aware of this interdependence: one player’s outcome
depends on their own and other players’ actions and all players know this.
We can contrast strategic with non-strategic situations in which the effect of
your actions on the outcomes you will experience is independent of what others do. Your enjoyment of the program you are streaming at home alone is
substantially independent of what others may be doing.
But many of our economic and social interactions are strategic:
G AME THEORY Game theory is the study
of strategic interactions using mathematical
models and verbal arguments to analyze
how the outcomes of the interaction for the
participants will depend on the rules of the
game and the objectives of the players.
E X A M P L E In 2020 under the pressure of
popular protests, the government of Chile
established a set of rules governing how
the constitution of Chile would be amended.
Another example of institutions is shown by
football (soccer). FIFA governs how football
can be played by what are called The Laws
of the Game. These institutions also change:
the corner kick was introduced in 1872 when
the U.K. Football Association changed the
rules.
S TRATEGIC I NTERACTION An interaction is
strategic when participants’ outcomes are
interdependent – their well-being depends
on the actions that both they and others
choose, and this interdependence is known
to the actors. An interaction is non-strategic
when this interdependence of people’s
outcomes is either absent or not recognized
by the participants. A short-hand expression
for the term strategic is: mutual dependence,
recognized.
• those considering driving to work know that their travel time will depend on
how others decided to get to work that morning;
• the Palanpur farmer knows that how his crop will fare if he plants early will
depend on how many others planted early
Checkpoint 1.4: Institutions
a. What are institutions?
b. What are "the rules of the game"?
Games
When we model strategic interactions using game theory we call the actors
players. Players can be people, firms, social movements, governments and
a variety of other entities. In biology, where game theory has been extensively used, even sub-individual entities are "players" such as viruses "trying
to" spread or genes "trying to" get as many copies of themselves made as
possible.
H I S TO RY John von Neumann (1903-1957)
was a Hungarian-American mathematician,
computer scientist, and physicist who is
regarded as the father of game theory,6
which he hoped would allow us to better
understand the anti-Semitism and fascist
political upheavals that he had witnessed in
the early 20th century and provide the basis
for understanding how groups interact.
32
MICROECONOMICS
- DRAFT
Players may choose from a list of possible strategies. For example, if private
property is an institution that is present and enforced, then a strategy set
might include "Purchase a Trek bicycle for $850." But it would not include
"Pick up any available Trek bicycle," without specifying the possible penalties
for stealing. The Palanpur farmers’ strategies are ‘Plant Early’ or ‘Plant Late.’
The strategies could also include a strategy based on what others did in the
past (called a contingent strategy) such as: "Plant early as long as at least 5
others planted early last season."
The description of a game requires us to identify the following:
• Players: a list of every player in the game whether they be individuals (like
the farmers in Palanpur), an organization such as Amazon or Alibaba, or
some other entity that can be represented as choosing between alternative
courses of action.
• Strategy sets: a list for each player of every course of action available to
them at each point where they must make a choice (including actions that
depend on the actions taken by other players, or on chance events). The
strategies selected by each of the players is called the strategy profile.
S ET A set in mathematics is a collection
of objects precisely defined either by
enumerating the objects, or by a rule for
deciding whether any particular object is
in the set or not. For example, the set of
positive, even integers less than or equal to
10 is, {2, 4, 6, 8, 10}.
• Order of play: a game can be simultaneous such that players make their
choices without knowing the choices of others, as in the game of rockpaper-scissors. Or a game can be sequential such that players move in
sequence, one after the other, as in chess, so that each player knows and
responds to the choices of the previous players.
• Information: A game also specifies
– who "knows" what,
– when do they "know" it,
E X A M P L E When we model the coordination
problem of the Palanpur farmers as a game
we assume they plant simultaneously. But
when we model the interaction between a
bank and a borrower we assume that the
banks first makes an offer (the loan size,
interest rate and schedule of repayment) and
the prospective borrower responds.
– is what they "know" known to others as well,
– can what they "know" be used in a court of law to enforce a contract,
and
– is what they "know" true (this is why we use the quotation marks)?
• Payoffs: Are numbers assigned to each possible outcome of the game
(each strategy profile) for each player; a player chooses a strategy with the
intention of bringing about the strategy profile with the highest number.
It is often useful to consider payoffs as something that the players actually
get. For example, considering the farmers in Palanpur again, an outcome
of the game is a strategy profile indicating who plants early and late, and
the payoffs could be the amount of grain each farmer harvests. We say that
the payoff associated with a particular outcome of a game is how much the
player values that outcome. But that means nothing more than that a player
C OOPERATIVE GAME A game in which
players can jointly agree upon how each
will play the game (and the agreement will
be respected or enforced) is a cooperative
game. If no binding agreement on how to
play the game is possible, then the game is
non cooperative.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
& ECONOMIC INSTITUTIONS
33
will choose a strategy resulting in an outcome with a higher payoff number if
possible.
An important distinction concerning strategy sets is whether or not one of the
strategies open to the players is to jointly agree on a strategy profile – that
is to deliberately coordinate their actions. This is possible in what is called a
cooperative game.
We use the set of players, their strategy sets, their payoffs, the order of play,
and the information the players have to describe the institutions governing
some economic interaction, whether it is between an employer and an employee, or a central bank like the U.S. Federal reserve and a commercial bank.
But even this detailed description of the interaction does not give us enough
information to predict how the game will be played.
The outcome of a game – how it will be played resulting in a particular strategy profile – also called a solution. To determine the solution as a way of predicting the outcome of a game we need what is called a solution concept. A
solution concept for a cooperative game would include some rule for deciding
on what the coordination would be, for example allowing one player selected
at random to dictate the outcome, or a particular system of voting.
But by positing some way that people could jointly implement some outcome,
S OLUTION CONCEPT A solution concept is
a rule for predicting the outcome of a game,
that is, how a game will be played.
cooperative game theory assumes away the problem of coordination. And the
problem of how coordination is to be achieved is at the heart of the classical
institutional challenge whether it takes the form of climate change or traffic
jams.
So we need to see how players might coordinate in what is initially a noncooperative setting – one in which coordination is not assumed at the outset
– lets take a concrete example. We will use this example to illustrate a basic
solution concept for non-cooperative games: the Nash equilibrium.
Checkpoint 1.5: Games
a. What is a game?
b. How do you describe the outcome of a game?
1.5 Over-exploiting nature: Illustrating the basics of game theory
People who fish for a living interact with each other regularly. Each of them
are aware that how much they benefit from fishing depends not only on their
own actions, but on the actions of others. This is because the more others
fish, the more difficult it will be for each to catch fish. The two fishermen
therefore impose negative external effects on each other, and this is why they
face a coordination problem. Given that they cannot jointly decide on how
Figure 1.4: Elinor Ostrom (1933-2012) was
an American Political Scientist who won the
Nobel Prize in economics for her work on social
dilemmas, such as those encountered by Alfredo
and Bob in the Fishermen’s Dilemma, and on
the institutions that promote cooperation in
groups. Photo Credit: Holger Motzkau. Wikimedia
Commons.
34
MICROECONOMICS
- DRAFT
Bob
12 Hours 10 Hours
Alfredo
10 Hours
12 Hours
Good
Worst
Best
Bad
much to fish, each faces a basic question: how much fishing to do given the
time are fishing the same waters?
The game set up
Specifically, we consider two fictional fishermen, Alfredo and Bob, who share
access to a lake, and catch fish, which they eat. There are no other people
affected by their actions.
Here we illustrate the basic concepts of game theory in a game we call the
Fishermen’s Dilemma. We chose the name because it is an example of what
is probably the most famous game, the Prisoners’ Dilemma. As before we use
a two-person example to illustrate a societal coordination problem among a
much larger number of actors.
The Fishermen’s Dilemma game is non-cooperative, which for two people
fishing in the same lake may seem unrealistic because as neighbors they
might be able to come to some kind of agreement about what each will do.
We do not consider this option in the two-person case because the model
illustrates a large number of people interacting. When many people interact
arriving at and enforcing such a cooperative agreement would present serious
challenges.
Here is the game.
• Players: Alfredo and Bob, two fishermen.
• Strategy sets: Each may fish for either 10 or 12 hours.
• Order of play : They simultaneously select a strategy, resulting in the
game’s strategy profile
• Payoffs: The players catch and eat an amount of fish given by the strategy
profile they have implemented.
This ends the game.
Figure 1.5: Alfredo’s payoffs to fishing more or
less depend on how much Bob fishes. Alfredo’s
payoffs are described using the words like we
used for the coordination problem: Planting in
Palanpur. Alfredo ranks his outcomes from best
to worst: Best > Good > Bad > Worst. Alfredo’s
strategies and outcomes are highlighted in Blue.
Bob’s strategies and outcomes are highlighted in
Red (but we have not put the words to describe
Bob’s outcomes in the figure).
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
2
(a) A’s payoffs only
35
Bob
10 Hours
12 Hours
3
1
4
2
(b) B’s payoffs only
Payoffs
The payoff of each player is composed of two parts:
• The amount of fish they are able to catch and consume, which they value
and would like to increase; and
12 Hours 10 Hours
4
1
12 Hours 10 Hours
3
10 Hours
12 Hours
Alfredo
12 Hours 10 Hours
Alfredo
10 Hours
Bob
Alfredo
Bob
& ECONOMIC INSTITUTIONS
12 Hours
3
3
4
1
1
4
2
2
(c) Payoffs for both players
Figure 1.6: Payoffs of players in the Fishermen’s
Dilemma. Alfredo’s payoffs are in the bottom-left
corner of each cell and are shaded blue. We
include Alfredo’s payoffs in the right-hand and
left-hand panels. Bob’s payoffs are in the top-right
corner of each cell and are shaded red. We include
Bob’s payoffs in the center panel and the left-hand
panel.
• The amount of time they spend fishing, which they find tiring and would like
to decrease.
We can describe the fishermen’s interaction in the form of a payoff matrix. (A
matrix is a rectangular array of quantities or other quantitative information).
We first present a version of the payoff matrix with words to represent Alfredo’s payoffs (but not yet Bob’s) in 1.5. Read the table this way: If Bob fishes
12 hours (the right hand column) and Alfredo fishes 10 hour (top row) this is
the worst outcome for Alfredo. A payoff matrix presents hypothetical ’if-then’
information; it presents all of the possible sets of payoffs, whether or not each
is likely ever to occur.
The complete payoff matrix for the Fishermen’s Dilemma is represented in
Figure 1.6 with numbers indicating the two fishermen’s evaluation of how
good the outcome indicated is. So for example the payoff to each if they both
fish ten hours (3) is fifty percent greater than if they both fish twelve hours
(2).
The convention we will use throughout this book is to list the row player’s
payoffs first and in the bottom left corner and the column player’s payoffs
second in the top right corner. So, in the Fishermen’s Dilemma game, we
list Alfredo’s payoffs first and Bob’s payoffs second. We shade each players
payoffs to make them easier to differentiate: blue for the row player (Alfredo)
and red for the column player (Bob).
Many of the games in this book involve two players and each player has two
possible strategies. We often call a game like this a “2 x 2” game (a “two-bytwo” game).
We now have all the elements we need for the complete description of the
N ORMAL F ORM G AME We will often describe
games using payoff matrices in what are
called normal or strategic form, like Figure
1.10. In normal or strategic form games,
we do not explicitly represent the time
sequence of the actions taken by each
player. We assume that each player moves
without knowing the move of the other
players. Normal form games therefore
often represent simultaneous move games,
games where players move at the same
time. Simple games in normal form are often
presented in a payoff matrix, a table that
includes all the relevant information about
the players, strategies and payoffs in the
game.
36
MICROECONOMICS
- DRAFT
Fishermen’s Dilemma and its strategy profiles and associated payoffs.
• Alfredo fishes 12 hours, Bob fishes 12 hours: When both fishermen fish
12 hours, they each catch fewer fish per hour of work, while they also have
a higher cost of effort because they’ve spent a lot of time fishing. Each
fisherman ends up with 2.
• Alfredo fishes 10 hours, Bob fishes 10 hours: When both fishermen spend
less time fishing they catch a decent amount of fish and they haven’t fished
so long that the other fisherman catches fewer fish. They also benefit from
a lower cost of time spent fishing. Each gets a net benefit of 3.
• Alfredo fishes 10 hours, Bob fishes 12 hours: Because Bob fishes 12
hours, Al catches many fewer fish and because Bob still fishes for another
two hours, he catches a lot of fish while Al doesn’t fish. Consequently, with
the cost of time and catching fewer fish, Al ends up with net benefits of 1
and Bob ends up with net benefits of 4.
• Alfredo fishes 12 hours, Bob fishes 10 hours: This is symmetrical to the
previous description, so now Al gets net benefits of 4 and Bob gets net
benefits of 1.
1.6 Predicting economic outcomes: The Nash equilibrium
As you already know, to predict a game outcome – the strategy profile that
will result– we need more than the description of the game alone. We need
to add what is termed a solution concept – a statement about how players
will behave in the game – that can be the basis of a prediction of the game’s
outcome. Predicting the outcome of a game – based on the rules of the game
and the solution concept – is especially important to understanding how
Figure 1.7: John F. Nash (1928-2015) was an
American mathematician who contributed the idea
of Nash equilibrium to game theory and won the
Nobel Prize in economics in 1994. His life was
documented in the book and movie A Beautiful
Mind.7 Source: Peter Badge, Wikimedia Commons.
changing the rules of a game can change the outcome of a game.
The key idea on which a solution concept is based is equilibrium. An equilibrium is a state in which there is nothing in the situation that will cause the state
to change. A predicted outcome will be an equilibrium, that is, an outcome
that is stationary (not changing). To understand why, imagine this were not the
case. You make a prediction, but then the outcome changes. Your prediction
would no longer be true because the outcome had changed.
Applying this reasoning to games, if we were to predict the outcome of a game
to be a strategy profile under which one or more players would have reason
to change their strategy, then the prediction would be falsified as soon as they
carried out the change. So the status of stationarity – change-less-ness – is a
property of a prediction; and this is why equilibrium is the fundamental idea of
making predictions about game outcomes.
Think of a concrete example. Suppose you want to predict where a marble
E QUILIBRIUM An equilibrium is situation that
is stationary (unchanging) because, as long
as the situation we are describing remains,
there is nothing causing it to change.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
& ECONOMIC INSTITUTIONS
37
will be if all that you know is that it is going to be somewhere in a round bottomed salad bowl sitting on a table. If I predicted that the marble would be
somewhere halfway up the side of the bowl you would doubt my prediction.
The reason is that any marble in that position would move downward in the
bowl, that is, its position would not be stationary so,if it ever were (for some
reason) where I predicted it would be, it would not be there any longer. It is
not that the prediction would necessarily be wrong. It could be true for a millisecond after I placed the marble in the bowl just above my predicted spot, for
example.
The only predicted position in the salad bowl that would not immediately falsify
itself in this sense is the bottom. So a reasonable prediction of the location of
the marble would be "the bottom of the bowl."
There are some situations in which a prediction based on an equilibrium
would be likely to be incorrect. Change the marble-in-bowl example by filling
the bowl with very thick honey. Then if you were asked to predict where the
marble would be found, you would want to know how long it had been in the
bowl, did have time to reach the bottom? If if the marble had been placed in
the bowl just a second ago, they you might be better off predicting that it would
be where it had been placed, rather than the bottom of the bowl.
The marble-in-bowl-of-honey is often a better illustration of how economic
processes work than the initial example. Markets are often out of equilibrium.
Predicting things in motion is a much more challenging task than predicting
them when they are stationary. We provide an example in our model of residential the segregation (below) where we are able to follow the process of
change step by step. But for the most part we study equilibria and how to
change them so as to improve outcomes.
In the marble-in-bowl illustration (without the honey) what is the solution concept that let us arrive the "bottom of the bowl" prediction? It is gravity, which
is our understanding about a reasonable way for the marble to "behave." In
modeling an economic interaction, the game structure is analogous to the
salad bowl. What is the analogy to gravity? The answer is the players’ best
response.
Best-response strategies
By far the most widely-used solution concept, the Nash equilibrium, is based
B EST RESPONSE A player’s best response is
a strategy that results in the highest payoff
given the strategies of the other players.
on the idea that players choose best-response strategies; they do the best
they can given the strategies adopted by everyone else
Bob
To understand better what a best response is, think about Alfredo’s choices in
10 Hours
sponses. First, what strategy should Alfredo adopt in order to gain the highest
Alfredo
for finding the Nash equilibrium on the basis of your analysis of his best re-
12 Hours 10 Hours
the Fisherman’s Dilemma. We will also introduce the "circle and dot" method
12 Hours
3
3
1
●
4
4
1
●
2
2
Figure 1.8: Hold constant Bob playing Fish 10
hours to assess Alfredo’s best response to Bob’s
playing Fish 12 hours.
38
MICROECONOMICS
- DRAFT
payoff if Bob were hypothetically to play Fish 10 hours (we say "holding constant" this strategy) as shown in Figure 1.8.
• Against Bob playing Fish 10 hours, Alfredo can get a payoff of 3 for fishing
10 hours or a payoff of 4 for Fishing 12 hours.
• 4 > 3 therefore fish 12 hours is Alfredo’s best response to Bob playing fish
Bob
10 hours.
10 Hours
Let’s repeat analysis and hold constant Bob playing Fish 12 hours, as shown
in Figure 1.9.
• Against Bob playing Fish 12 hours, Alfredo can get a payoff of 1 for playing
Fish 10 hours or a payoff of 2 for playing Fish 12 hours.
12 Hours 10 Hours
10 hours) to indicate that it is Alfredo’s best response.
Alfredo
• place a solid point in the cell (Alfredo plays Fish 12 Hours, if Bob plays Fish
12 Hours
3
3
●
4
4
1
1
●
2
2
Figure 1.9: Hold constant Bob playing Fish 12
hours to assess Alfredo’s best response.
• 2 > 1 therefore Fish 12 hours is Alfredo’s best response to Bob playing fish
10 hours.
• place a solid point in the cell (Alfredo plays Fish 12 Hours, Bob plays Fish
12 hours) to indicate that it is Alfredo’s best response.
Checkpoint 1.6: A best response for Bob
Repeat the process we went through for Alfredo, but do it for Bob instead. Notice that when you do so, you will blank out a row for Alfredo to hold his strategy
constant, whereas you blanked out a column for Bob to hold his strategy constant. What are Bob’s best responses? Show his best responses using a hollow
circle.
M CHECK: STRONG AND WEAK BEST
R E S P O N S E . A best response may be either
strong or weak. A strong (also called strict)
best response yields higher payoffs than
any other: it is strictly "better" than any other
strategy. There can be no strategy that is
better than a weak best response but a weak
best response need not be better than any
other; it may be "as good as" (the payoffs to
the strategy and some alternative strategy
being equal.)
Nash Equilibrium and the outcome of a game
Using the best responses of the players we can now predict the outcome
of a game using as our solution concept the Nash equilibrium. A Nash
equilibrium is a profile of strategies – one for each player – each of which is
a best response to the strategies of the other players. A Nash equilibrium is
also called a mutual best response. Because at a Nash equilibrium all players
are playing their best response to all of the others, it follows that no player
has a reason to change his or her strategy as long as the other players do not
change theirs. Some games do not have a Nash equilibrium and you will see
shortly that some have more than one.
In Figure 1.10, Alfredo’s best responses are shown by the solid black dot
in the cell. Bob’s best responses are shown by the hollow circle. Their best
responses coincide at the Nash equilibrium (Fish 12 Hours, Fish 12 Hours)
with payoffs (2, 2) shown in the cell where the solid point is inside the hollow
circle. You can use the "dot and circle" method to find one or more Nash
E X A M P L E The Rock,Paper, Scissors game
(also called ro-sham-bo and by many other
names in other languages) originated in
China about two thousand years ago. It does
not have a Nash equilibrium.
N ASH EQUILIBRUM A Nash equilibrium is a
profile of strategies – one strategy for each
player – each of which is a best response to
the strategies of the other players.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
12 Hours 10 Hours
Alfredo
12 Hours
3
3
4
4
1
●
1
39
Figure 1.10: Payoff Matrix for The Fishermen’s
Dilemma. The solid dots indicate Alfredo’s best
responses. The hollow circles indicate Bob’s
best responses. A Nash equilibrium is a cell that
contains both dots. In this case there is just one
Nash equilibrium: both fishing 12 hours.
Bob
10 Hours
& ECONOMIC INSTITUTIONS
●
2
2
equilibria (if they exist) for games that can be represented by a payoff matrix
like Figure 1.10.
In the Fishermen’s Dilemma game described by Figure 1.10. Each player’s
best response to both of the others’ strategies was Fish 12 hours. Therefore
only one outcome is a Nash equilibrium: both fishermen fishing 12 hours. We
say that (Fish 12 Hours, Fish 12 Hours) is the Nash equilibrium with payoffs
(2, 2).
The outcome demonstrates how Nash equilibrium can initially seem counterintuitive. Both would have had higher payoffs if they could have agreed to
restrict their fishing to 10 hours (they could have had 3 each if they both fished
10 hours and 3 > 2). But suppose both were for restricting their fishing to
10 hours; then both would an incentive to fish for 12 hours (because 4 > 3)
and unless they had a binding agreement to continue fishing less, both would
choose to fish more.
The Fishermen’s Dilemma is therefore a coordination problem and it returns
us to the classical institutional challenge. Without institutions to align the
individual interest of the participants with their shared interest, they get an
outcome that is worse for both of them than other possible outcomes. We
will later show how a change in the institutions reglating how Alfredo and
Bob interact – that is, changing the rules of the game – might address this
coordination.
Checkpoint 1.7: Nash Equilibrium
a. Explain why none of the other three outcomes (those that are not, (Fish 12
Hours, Fish 12 Hour) of the Fishermen’s Dilemma satisfy the definition of
Nash equilibrium.
b. At each of the other three outcomes, which player has an incentive to
change strategy and in what way? Explain.
c. Explain why a game like Rock Paper Scissors would not be much fun if there
was a Nash equilibrium.
40
MICROECONOMICS
- DRAFT
Dominant strategies
In the Fisherman’s Dilemma (and all Prisoners’ Dilemmas) there is a single
strategy that yields the highest payoffs to a player for both (or all of if there
are more than two) of the strategies that the other player might adopt. A
strategy is a player’s dominant strategy if it is the player’s best response
to all possible strategy profiles of the other players. That is, a strategy is a
dominant if by playing it the player’s payoff is greater than or equal to the
payoff playing any other strategy for every one of the other player’s profiles of
strategies.
Likewise we say that strategy A is dominated by another strategy B if the
payoff to playing B is at least as great or greater than playing A for every
strategy profile of the other players. If there is a strategy that dominates all of
the other strategies that an player may choose, then it is a dominant strategy.
If each player in a game has a dominant strategy, then the strategy profile in
D OMINANT STRATEGY A strategy is dominant
if by playing it the player’s payoff is greater
than or equal to the payoff playing any
other strategy for every one of the other
players profiles of strategies. A strategy is
dominant if it is the player’s best response
to all possible strategy profiles of the other
players. A dominant strategy dominates all of
the other strategies available to the player.
which all players adopt their dominant strategy is called a dominant strategy
equilibrium.
We can apply the concept of dominant strategy equilibrium to the Fishermen’s Dilemma. To do so, we need to understand whether each player has a
dominant strategy.
• When Alfredo fishes 10 hours, his payoff is 3 if Bob fishes 10 hours and 1 if
Bob fishes 12 hours.
• When Alfredo fishes 12 hours, his payoff payoff is 4 when Bob fishes 10
hours and 2 when Bob fishes 12 hours.
• So, when Bob fishes 10 hours, fishing 12 hours gets Alfredo a higher payoff
(4 > 3) and when Bob fishes 12 hours, fishing 12 hours gets Alfredo a
higher payoff (2 > 1)
• Therefore, Alfredo gets a higher payoff from fishing 12 hours against each
of his opponent’s strategies
• Fish 12 hours is therefore Alfredo’s dominant strategy.
Fishing 12 hours is also Bob’s dominant strategy. Because each player has a
dominant strategy to fish 12 hours, the dominant strategy equilibrium is (Fish
12 hours, Fish 12 hours) with payoffs (2, 2).
The fact that the Fishermen’s Dilemma has a dominant strategy equilibrium
makes it a particularly simple problem (both for us, studying it, and for the
players because what is best for each does not depend on what the other
does). The dominant strategy equilibrium of a game is always a Nash equilibrium, so in the Fishermen’s Dilemma, the outcome where both players fish 12
hours is the only Nash equilibrium.
D OMINANT STRATEGY EQUILIBRIUM . A
dominant strategy equilibrium is a strategy
profile in which all players play a dominant
strategy.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
Checkpoint 1.8: Dominance and Nash Equilibrium
a. Repeat the analysis we did for Alfredo for Bob and confirm that 12 hours is a
dominant strategy for him too.
b. We said that a dominant strategy equilibrium is always a Nash equilibrium.
But do you think that a Nash equilibrium is always a Dominant Strategy
equilibrium? Why or what not?
1.7 Evaluating outcomes: Pareto-comparisons and Pareto-efficiency
The Nash equilibrium can help us predict the result of a particular interaction.
But it does not tell us anything about whether some outcome is good by any
standard, or even better or worse than some other outcome. Economists,
policy-makers and others would like to evaluate whether some outcomes are
better or worse so that we can try to work out which rules of the game would
make the better outcomes Nash equilibria, and therefore more likely to be
what we observe. In Chapter 16 we show how economics deals with this for
questions of public policy.
The challenge in making these comparisons is that whether some outcome is
better than another depends on what you value, and there is no agreed upon
standard of what makes one outcome better than another. Returning to our
fishermen, here are some of the values that one could use in evaluating an
outcome
• Fairness in the distribution of payoffs among the players; is it fair that
Alfredo receives 4 times what Bob gets when Alfredo does not limit his
fishing hours and Bob does?
• Are the rules of the game itself fair? In the Fishermen’s Dilemma the same
rules applied to both players; but were the game a bit different, many would
think it unfair if Alfredo could simply order Bob to fish 10 hours, or to hand
over half of all the fish Bob caught.
• Setting aside fairness, is the outcome a reasonable use of available resources including the working time of the two fishermen and the sustainability of the lake itself and the living things that it supports.
There are many other standards that could be proposed. Questions of better
and worse are called "normative question." With normative questions matters
of ethics or morals are necessarily involved. We will introduce some experimental evidence on people’s views concerning fairness in Chapter 2 and
some analytical tools for studying normative issues in Chapter 13.
Here we introduce a concept that economists use to evaluate economic outcomes. The idea is simple: an objective of public policy and institutional design – the rules of the game – should be to avoid those outcomes – like traffic
& ECONOMIC INSTITUTIONS
41
42
MICROECONOMICS
- DRAFT
6
Bob
12 Hours 10 Hours
Alfredo
10 Hours
12 Hours
3
3
c
b
4
d
3
c
2
a
4
1
1
4
Bob's payoffs
5
d
2
2
a
1
b
0
0
(a) Fishermen’s Dilemma with Labeled Points
1
3
4
5
6
Alfredo's payoffs
(b) Analyzing points for Pareto efficiency
jams, planting late in Palanpur, and over-fishing the lake – that are worse for
everyone, compared to an alternative outcome that also would have been
feasible.
Pareto comparisons
To compare outcomes when more than one person is involved economists
use the concept of Pareto-efficiency based on Pareto-comparisons of
outcomes.
An outcome is Pareto superior to another if it allows at least one of those
involved to be better off without anyone being worse off. A Pareto-superior
outcome is also called a Pareto improvement over the outcome it was compared to. This is a Pareto comparison. An outcome is Pareto efficient if no
other feasible outcome is Pareto superior to it.
Figure 1.11 depicts the outcomes of the interaction between Alfredo and
Bob. Alfredo’s payoffs are on the horizontal axis (x-axis), so the outcomes get
better for Alfredo as you move from the left to the right. Bob’s payoffs are on
the vertical axis (y-axis), so the outcomes get better for Bob as you move from
bottom to top. The left panel of Figure 1.11 is the Fishermen’s Dilemma payoff
matrix with each outcome given a label a, b, c, or d. These same points
appear in the right panel of Figure 1.11 where you can read on the vertical
and horizontal axes the payoffs to the two players that you see in the payoff
matrix.
The Pareto-comparison is geometrically easy to see in this type of plot. An
outcome is Pareto-superior to another if the first outcome lies to the "northeast" of the second in the plot. "North-east" in this figure is "better for both."
An outcome is Pareto-efficient if there is no other feasible outcome to the
2
Figure 1.11: Three Pareto-efficient outcomes
of the Fishermen’s Dilemma. In 1.11 a, the
outcomes are labeled in the bottom right corner
as a for (2, 2), b for (4, 1), c for (3, 3) and d for (1,
4). These allow us to make Pareto comparisons
for the outcomes of the Fishermens Dilemma.
Alfredo’s payoffs are plotted on the horizontal axis,
increasing as you move rightward. Bob’s payoffs
are plotted on the vertical axis, increasing as you
move upward.
In 1.11 b we show the Fishermen’s Dilemma
indicated by the four possible outcomes given by
the same letters that appear in each of the cells of
the payoff matrix. We use shaded colors indicating
90 degree angles to the northeast of the feasible
outcomes (each of the lettered points).
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
& ECONOMIC INSTITUTIONS
43
north-east. Or if you think of the colored areas whose lower left corners are
points a, b, c, and d then a Pareto efficient point is one that has no other point
in its "colored shadow" extending upwards and to the right of the point.
When two different outcomes are both Pareto-efficient, they cannot be Paretocompared or Pareto-ranked. We could rank c above a because both players
were better off, but with b, c and d we cannot move from one outcome to
another without worsening outcomes for at least one of the players.
PARETO
PARETO
An outcome is Pareto superior
to another if it allows at least one of those
involved to be better off without anyone
being worse off. A Pareto-superior outcome
is also called a Pareto improvement over the
outcome it was compared to. This is a Pareto
comparison. An outcome is Pareto efficient if
no other feasible outcome is Pareto superior
to it. If we can rank two outcomes such that
one is Pareto superior to the other, then we
say that these two outcomes can be Pareto
compared, or Pareto ranked.
COMPARISONS AND
EFFICIENCY
Here is a checklist to use when evaluating any given set of payoffs for both
players. Consider a point called x with payoffs pxA for Player A and pxB for
player B and compare it to another theoretical point, y with its corresponding
payoffs:
• For any point, x, check whether there exists an alternative point y where
at least one player gets a payoff that is greater than pxA or pxB without the
other player being worse off.
• If at y, pyA > pxA while pyB pxB or pyB > pxB while pyA pxB then y is Paretosuperior to x (at least one player is better off while the other player is not
worse off).
• If no other point exists where at least one player is better off with no other
player being worse off (no Pareto-superior point exists), then x is Paretoefficient.
Checkpoint 1.9: Pareto efficiency and Pareto improvements in the Fishermen’s Dilemma
Referring to Figure 1.11, consider the following:
a. Is any point dominated by some other point? Say which, if any? item At
which point is the total payoff of the two fishermen the greatest?
b. Would a change from any other point to that "total payoff maximum" point be
a Pareto improvement?
1.8 Strengths and shortcomings of Pareto efficiency as an evaluation of outcomes
Pareto efficiency gives us a way to identify "lose-lose" outcomes we should
seek to avoid, namely those " that are worse for all than they could be" . But,
as we will now see, except in special cases, Pareto efficiency does not provide
a rule to select what we might call in everyday speech "the best" outcome. To
see why this is true, suppose we have a cake of given size and we are dividing
it among people, all of whom equally enjoy eating cake.
An outcome in which one person gets the entire cake is surely Pareto-efficient
because in any other allocation that lucky person would get less. Likewise an
E X A M P L E We will see in Chapter 2 that
people often reject a highly unequal division
of some "pie" and are willing to sacrifice a
substantial amount of money rather than to
accept what they consider to be an unfair
division. And we will see in Chapter 13 how
people’s desires for a fair society might allow
for a choice among outcomes that differ in
who gets what.
44
MICROECONOMICS
- DRAFT
allocation in which everyone got an equal slice of the cake is Pareto-efficient,
for in any other allocation at least one person would have to get less. The
Pareto criterion can say nothing about such distributional fairness. All it says is
"make sure there’s no cake left on the table!"
When there are many Pareto-efficient outcomes there is always a conflict of
interest among players over which outcome they would prefer we cannot say
that one is "more Pareto-efficient" than the other.
It is also perfectly sensible to prefer an outcome that is not Pareto efficient but
is more fair over an alternative Pareto efficient outcome that is unfair. To continue the cake example, if there are two people between whom the cake will
be divided many people would reject the (Pareto efficient) outcome in which
one person gets the entire cake in favor of a Pareto inefficient alternative in
which each gets a third of the cake (the remaining third perhaps being thrown
away or destroyed in the conflict over its distribution). But the Pareto comparison would remind us that each person getting half of the cake is preferable to
each getting a third with the rest being wasted.
Pareto efficiency is a particular device for screening out those outcomes (like
throwing away some of the cake in the above example, or planting late in
Palanpur) that should not be among the list of candidate feasible outcomes
among which the choice of better or best should be made or grounds of fairness or other bases.
Other ways of "screening" the list of candidate outcomes would give priority
not to individuals’ payoffs but to whether the rules of the games that produced
the outcomes are themselves fair and consistent with other values such as
individual dignity, respect and freedom.
Checkpoint 1.10: Pareto efficiency
Consider these questions about Pareto efficiency.
a. True or False (and explain): "The fact that an outcome is Pareto-efficient
does not imply that it is preferred by all the actors to all the other outcomes."
b. Can two Pareto efficient outcomes be Pareto compared? Why or why not?
Explain.
c. Imagine you are an impartial observer evaluating the possible outcomes that
might occur for Bob and Alfredo. Are there any reasons why you might judge
the Pareto-inefficient outcome a in the figure to be better than the Pareto
efficient outcomes b and d, despite the fact that a is Pareto-inefficient?
1.9 Conflict and common interest in a Prisoners’ Dilemma
The game the fishermen are playing is a particular case of the Prisoners’
Dilemma. A Prisoners’ Dilemma is a two-person interaction in which there is
P RISONERS ’ D ILEMMA A Prisoners’ Dilemma
is a two-person social interaction in which
there is a unique Nash equilibrium (that is
also a dominant strategy equilibrium), but
there is another outcome that gives a higher
payoff to both players, so that the Nash
equilibrium is not Pareto-efficient.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
10 Hours
(Defect) (Cooperate)
12 Hours
Alfredo
12 Hours
(Defect)
y
w
w
x
x
y
45
Figure 1.12: A General Prisoners’ Dilemma.
For the game to be a Prisoners’ Dilemma, we
require y > w > z > x and 2w > y + x (this is like
4 > 3 > 2 > 1 and 2 ⇥ 3 > 4 + 1 from the numerical
example).
Bob
10 Hours
(Cooperate)
& ECONOMIC INSTITUTIONS
z
z
a unique Nash equilibrium (that is also a dominant strategy equilibrium), but
there is another outcome that gives a higher payoff to both players, so that
the Nash equilibrium is not Pareto-efficient. So, in the Prisoners’ Dilemma
both players get their second worst payoffs in the game by playing their strictly
dominant best-response strategies.
We now point out some of the general characteristics of this particular kind of
coordination problem. To do this, in Figure 1.12 we show the familiar payoff
matrix for the Fishermen’s Dilemma, but instead of the numbers indicating
the payoffs of the players now we label the payoffs w, x, y, and z. We label
the action of fishing 10 hours "Cooperate" because it is the action the two
fishermen would take if they could coordinate their actions. The action of
fishing 12 hours is labeled "Defect" because a fisherman who chooses to fish
12 hours is deviating from the mutual cooperate outcome on which the two
fishermen might be able to coordinate.
The interaction is a Prisoners’ Dilemma if two conditions hold:
• y > w and z > x means that fishing Defect is a strict dominant strategy
• w > z means that mutual cooperation is Pareto superior to mutual defection.
For Alfredo, 12 Hours is a best response to Bob playing 10 Hours because
y > w; 12 Hours is also a best response to 12 Hours because z > x (both best
responses are shown by the solid point). Similarly, for Bob, 12 Hours is a best
response to Alfredo playing 10 Hours because y > w; 12 Hours is also a best
response to 12 Hours because z > x (both best responses are shown by the
hollow circle). Therefore the Nash equilibrium is (12 Hours, 12 Hours) with
payoffs (z, z).
If the players play the game non-cooperatively (they do not coordinate their
actions) each will play their dominant strategy – defect – and get z when by
cooperating they could have each received w.
M AT H N OT E A third condition is sometimes
added, namely x + y < 2w which means
that the sum of payoffs when both players
cooperate is greater than the sum of payoffs
when one cooperates and the other defects.
This condition makes (Cooperate, Cooperate) preferable to any outcome in which one
defects and the other cooperates.
46
MICROECONOMICS
- DRAFT
Economic rent: The incentive to coordinate
Both players have a good reason to try to change the rules of the game so
that they can agree on both cooperating. How much more they would get if
they were to mutually cooperate than if they mutually defected – in this case
w
z – is called an economic rent, meaning the difference between the
payoff that they would get if they cooperated and their next best alternative.
Their next best alternative to cooperating, we assume, is mutual defection,
also known as their fallback position.
FALLBACK POSITION A player’s fallback
position is the payoff they receive in their
next best alternative.
Economic rents and the fallback position play a central role in microeconomic
theory, so it is a good idea to master them. The meaning of fallback position
is intuitive: it is what you fall back to if your current outcome is not possible,
in this case if the mutual cooperation should not work out. A player’s fallback
position is the payoff they receive in their next best alternative.
The term "economic rent" may at first seem surprising, because the word
"rent" also means a payment for the temporary use of something like a rent
paid to a landlord or a rent or a car rental agency. The term economic rent
means something entirely different. A participant’s economic rent is the payoff
they receive in excess of what they would get in their fallback position.
We shall use the idea of a fallback often, from social interactions like the
Prisoners’ Dilemma, to worker-employer relationships where a worker wants
a job more than being unemployed, to a person applying to a bank for a loan
rather than trying to get money from friends, family, or the government. As
these examples indicate, in real life the fallback position will differ depending
E CONOMIC RENT A participant’s economic
rent is the payoff they receive in excess
of what they would get in their fallback
position. When we use the term " rent"
we mean economic rent. The sum of the
economic rents received by the participants
in an interaction is sometimes termed the
economic surplus.
on the details of the situation that we set aside when we model interactions
like fishing in a lake, employment and borrowing.
Impediments to coordination: Limited information and conflicts of interest
If w
z is substantial – meaning substantial rents associated with cooperation
for each player – then it might seem a simple matter for the players to agree
to cooperate. But people often fail to reach or enforce such an agreement, for
two main reasons:
• Limited information. The participants may lack the information needed to
monitor and enforce an agreement. How can a participant know or verify
what other participants do?
• Conflict over distribution of the economic rents from cooperation Disagreement about who gets what – for example who gets to fish more – may
make it impossible for the two to agree.
Concerning the information problem, the fishermen, for example, may have no
way of enforcing an agreement, or even knowing if the agreement has been
violated. While each may know how many hours the other has fished on day
E X A M P L E The term economic rent is what
is known in the study of language as a
“false friend,” a term that you think you
know the meaning of but mean something
entirely different in the new language you
are now learning. "Sensible" in English
means "reasonable" but in Italian it means
"sensitive."
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
& ECONOMIC INSTITUTIONS
with clear and sunny weather, on a foggy day it may be impossible to know.
Even if one fisherman knows how much the other fished, that knowledge may
be insufficient to enforce an agreement through a third party such as a court
of law.
This is the problem of asymmetric information or non-verifiable information. Information is asymmetric if people know different things, or if what one
person knows (for example how many hours he fished), the other person does
not know. Information is not verifiable if people cannot use it to enforce an
agreement or a contract. For example most courts will not accept "hearsay"
(meaning "second hand") information, so if one of the fishermen had heard
from someone else that the other had fished 12 hours, this would be nonverifiable information.
Asymmetric and non-verifiable information will play a central role in our analysis of how the labor market, the credit market and other markets work.
Concerning the conflicts over the distribution of the economic rents from
cooperation, the Fishermen’s Dilemma, the agreement to restrict fishing to 10
hours a day, is an agreement both to restrict fishing and to divide the benefits
of restricting fishing in a particular way, namely equally. But the fishermen
need not agree on 10 hours each. Alfredo might insist that he will fish 12
hours and Bob only 10 hours. Or Bob might insist on the opposite.
Or Bob might insist that both fish 10 hours, but that Alfredo give him most of
Alfredo’s catch, leaving Alfredo with just enough of his catch to be no worse
off than had they both fished 12 hours, namely with a payoff of z. Unless they
can find a mutually acceptable solution to the distribution problem they may
end up having no agreement at all, and then simply fish at 12 hours each, at
their fallback position.
The fishermen’s distribution conflict highlights a challenge that arises in any
voluntary economic interaction. Consider their possible agreement to limit
their fishing time:
• The agreement is voluntarily entered into. This means that neither player
can force the other to accept terms worse than their fallback position.
• The agreement therefore must allow each participant to achieve a payoff
greater than (or at least not worse than) had the individual not agreed
to cooperate. In other words, there must be some economic rents made
possible by a voluntary cooperative outcome.
• This being the case, the participants have to find a way that the total rents
will be divided.. If they are to agree to cooperate by restricting fishing, they
must also agree on how these economic rents will be distributed.
• Conflict over the distribution of the economic rents (who gets what amount
ASYMMETRIC INFORMATION Information is
asymmetric if something that is known by
one participant is not known by another.
N ON - VERIFIABLE INFORMATION cannot
be used to enforce a contract or other
agreement
47
48
MICROECONOMICS
- DRAFT
of economic rent) may prevent the fishermen from coming to an agreement.
We sometimes think of cooperation and conflict as opposites, as for example when members of a team cooperate in their efforts to win some conflict
with another team. But the Prisoners’ Dilemma is a scenario of conflict and
cooperation among the very same participants. They have common interests
in getting some share of the economic rents by cooperating; but they have
conflicting interests in how the total will be divided into the rents received by
each.
A catalogue of games: And their challenges to coordination
Some interactions present greater impediments to coordination than other; the
Prisoners’ Dilemma is in some respects the most challenging of all.
We can classify coordination problems and the challenges they present by the
relation between Nash equilibria and Pareto-efficient outcomes of the games
that represent them.
• In the Prisoners’ Dilemma, you know, there is a unique Nash equilibrium
that is Pareto-inefficient. Because this outcome is also a dominant strategy
equilibrium, coordination on mutual cooperation will not occur even if one of
the players insists (perhaps for moral reasons) on cooperating.
• In interactions like Planting in Palanpur, which are often called Assurance
Games, there are two Nash equilibria, (both Plant Early and both Plant
Late) one of which (Plant Early) is Pareto-superior to the other (Plant Late).
In these games if one of the players plays the strategy making up the
Pareto superior equilibrum (Plant Early) then the best response of the other
will be to do the same. Finding institutions that will implement the preferred
plant early outcome in a game like this will be a lot less challenging than in
a Prisoners Dilemma.
• Another important class of coordination problems arise in what we call
Disagreement Games where there are two Nash equilibria each of which
is Pareto-efficient, so that they cannot be Pareto-ranked, and players disagree about which Nash equilibrium they would like occur. These are like
the Planting in Palanpur game but with the additional challenge stemming
from a conflict over which Nash equilibrium will be implemented.
We start with an even less challenging game in which players’ self interests
lead them to a Pareto-efficient Nash equilibrium.
F AC T C H E C K In the next chapter we will
see that across many cultures of the world,
people would rather get nothing than get
what they consider to be an unfair share
of the economic rents, and as a result
cooperation breaks down and nobody gets
any rent at all.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
Barbara
Tomatoes
1.10
Corn
Tomatoes
Arkady
Corn
2.5
1
49
Figure 1.13: An Invisible Hand Game with best
responses indicated by circles and dots. Arkady’s
payoffs are listed first in the bottom-left corner.
Barbara’s are listed second in the top-right corner.
The game captures Adam Smith’s ideas of
specialization and gains from trade (that is, the
opportunity to obtain economic rents from trade).
2
2
●
4
& ECONOMIC INSTITUTIONS
4
●
1
2.5
Coordination successes: An invisible hand game
The characteristic of an invisible hand game is that it has a Nash equilibrium
that is Pareto-efficient. The Invisible Hand game illustrates Adam Smith’s
core insight that through the competitive buying and selling of privately owned
goods on competitive markets, self-interested people can achieve outcomes
to the benefit of all at least some conditions (that we spell out in Chapter 14).
In modern economic language, we would say they avoid Pareto-inefficient
outcomes.
Though our game is much simpler than Smith’s reasoning and Smith did
not use ideas like Pareto efficiency, our game illustrates an aspect of Adam
Smith’s idea of how the invisible hand works. The participants, pursuing their
self-interest, reach an outcome that beneficial for all of them. (We return to
how the invisible hand is understood in contemporary economics in Chapter
14).
Consider a 2-by-2 game between two players, Arkady and Barbara, who are
both farmers. Each player can choose between two strategies: planting corn
or planting tomatoes. The payoffs that they achieve are provided in the payoff
matrix in Figure 1.13, which we call the Corn-Tomatoes game.
The payoff matrix reflects two facts about the problem that the two farmers
face.
• Either because of their skills or the nature of the land they own, Arkady is
better at growing tomatoes; Barbara is better at growing corn
• They both do poorly when they produce the same crop because the increased supply of whichever good it is that they both produce drives down
the price.
The Nash equilibrium of the Corn-Tomatoes game is (Tomatoes, Corn), that is,
Arkady plants tomatoes, and Barbara plants corn, at which the players receive
payoffs (4,4). (Tomatoes, Corn) is Pareto efficient as there is no alternative
I NVISIBLE H AND G AME In an invisible hand
game there is a Nash equilibrium that is
Pareto-efficient.
50
MICROECONOMICS
- DRAFT
5
Bina
Late
3
4
4
c
0
Pareto−efficient
Nash equilibrium
c
4
Bina's payoffs
Early
Late
Aram
Early
A's fallback
b
0
2
d
a
3
b
a
2
Pareto−inferior
Nash equilibrium
1
d
0
3
2
0
(a) Assurance Game with labeled payoffs
1
3
4
5
(b) Payoffs plotted against each other
Just as in Adam Smith’s reasoning about his invisible hand, Arkady and Barbara, in their interaction, through competition with each other and following
their self-interest coordinate their economy to their mutual benefit. In the Invisible Hand Game, each player pursues self-interested objectives and benefits
from the fact that the other does too. In an Invisible Hand game individual
incentives lead people to act in ways that promotes mutual benefit.
Checkpoint 1.11: Invisible Hand Game
Which entries in the payoff matrix would you have to compare in order to show
the follwing:
a. They each do better when Arkady specializes in tomatoes and Barbara
specializes in corn then vice versa.
b. They each do worse when both produce the same crop.
c. Growing corn is Barbara’s dominant strategy
d. Arkady growing tomatoes and Barbara growing corn is the dominant strategy
equilibrium.
e. Explain why the Nash equilibrium of the game is Pareto efficient.
Assurance Games: Win-win and lose-lose equilibria
Return to the farmers in Palanpur. There are two Nash equilibria in this game,
one in which both participants Plant Early and one in which both Plant Late.
The best response to the other farmer planting early is also to plant early,
while the best response to the other farmer planting late is also to plant late.
The outcome where both farmers plant early is Pareto-superior to the outcome
when both farmers plant late.
2
Aram's payoffs
outcome which is Pareto superior to it.
1.11
B's fallback
Figure 1.14: Planting in Palanpur: An Assurance
Game. Aram’s payoffs are listed first in the
bottom-left corner. Bina’s payoffs are listed second
in the top-right corner. Aram’s best responses are
shown by the solid point and Bina’s are shown
by the hollow circle. The Nash equilibria of the
game are (Plant Early, Plant Early) and (Plant
Late, Plant Late), with payoffs (4, 4) and (2, 2). The
Plant Early Nash equilibrium is Pareto-efficient.
The Plant Late equilibrium is not. In the right-hand
panel, the payoffs are plotted against each other.
Aram’s payoffs are plotted on the horizontal axis,
increasing as you move rightward. Bina’s payoffs
are plotted on the vertical axis, increasing as you
move upward.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
& ECONOMIC INSTITUTIONS
51
The players do not have any conflict of interest: both would share equally
in the gains to cooperation, should they find a way to coordinate on planting early. The problem for the real life farmers of that village is that they are
stuck in the Pareto-inefficient Nash equilibrium of an Assurance Game. Their
challenge is how move to the Pareto-superior Nash equilibrium.
This could happen if all the participants had confidence (were assured)
that the other participants would follow their lead in moving to the superior
outcome. This is why this type of game is often labeled as an "Assurance
Game."
Figure 1.14 is the payoff matrix for two players, Aram and Bina, choosing
A SSURANCE G AME In an Assurance Game,
there are two Nash equilibria, one of which is
Pareto-superior to the other. The Planting in
Palanpur Game is an example.
when to plant their millet in the village of Palanpur, India. (It is the same as
the earlier figure about the two farmers, except that we now have numbers
representing the farmers’ payoffs). Coordination failures arise in the Assurance Game because of positive feedbacks: the more people who plant late
the more is the incentive for others to plant late, and vice versa. Each strategy
exhibits strategic complementarity.
Checkpoint 1.12: Graphing Palanpur
Using the graphical method for identifying Pareto-efficient outcomes as shown
in Figure 1.14, show which outcomes in the Palanpur game are Pareto-efficient.
Can you explain why a and c are Nash equilibria?
Assurance game and strategic complementarity
Social media, dating platforms, and other matching services are examples of
strategic complementarities. They are more valuable to for everyone if many
people participate.
Strategic complementarity exists when either of two conditions hold.
1. A strategy is a strategic complement to itself : The payoff to playing a
particular strategy increases as more people adopt that strategy as a result
of some form of positive feedbacks. Dating platforms are an example. The
strategy could be “Open a Tinder Account.” The positive feedback arises
because the more other people that are using Tinder the more people you
will “meet.” Plant Early in Palanpur is another example as we will see.
2. One strategy and another are strategic complements to each other. The
payoff to playing one strategy (say, A) is greater the more people adopt
the other (B) in which case we say that strategies A and B are strategic
complements. An example is the Invisible Hand Game shown in Figure
1.13. The payoff to Arkady from planting tomatoes is greater if Barbara
plants corn (instead of tomatoes), and the payoff to Barbara from planting
S TRATEGIC C OMPLEMENTARITY Strategic
complementarity exists when i) A strategy is
a strategic complement to itself : The payoff
to playing a particular strategy increases
as more people adopt that strategy as a
result of some form of positive feedbacks, or
ii) One strategy and another are strategic
complements to each other. The payoff to
playing one strategy (say, A) is greater the
more people adopt the other (B) in which
case we say that strategies A and B are
strategic complements.
52
MICROECONOMICS
- DRAFT
corn is greater if Arkady plants potatoes (instead of corn). Growing corn
and growing tomatoes are strategic complements.
The farming in Palanpur problem is a case of the first, not the second. But
nobody has any reason to participate if no others do. This is an example of
what are called network externalities or network external effects which
occur when the benefits to members of a social or physical network increase
when more people join the network.
One example is a particular social network: if you’re the only person on it,
there is really no point. But, as more and more people join the network,
then social networking site becomes more useful and your payoff increases
with the number of users in the network. These so-called network externalities are a particular case of strategic complementarity. By joining a network
each person confers an external benefit on the existing members of the network.
We predict and evaluate the possible outcomes of the Planting in Palanpur
Game using the concept of best response (using the dot and circle method
introduced earlier). And we see that the game has two Nash equilibria (Early,
Early) with payoffs (4, 4) and (Late, Late) with payoffs (2, 2). (Early, Early) is
Pareto-superior to (Late, Late) and it is Pareto-efficient because no alternative
outcome is Pareto-superior to (Early, Early).
Even though there is a Pareto efficient Nash equilibrium, a population – like
the people of Palanpur – may get stuck in the Pareto inferior Nash equilibrium. that does not guarantee players will actually play it. So we have two
conclusions:
• The fact that a Pareto efficient outcome is a Nash equilibrium does not
mean that it will be the one we observe and
• In cases where there is more than one Nash equilbrium, we need more
information than is provided by the Nash equilibrium and Pareto efficiency
concepts to make a prediction about the strategy profiles we will see in
practice.
Checkpoint 1.13: Assurance Game
Which payoff table entries would you have to compare in order to show that:
a. Planting early is Pareto efficient.
b. Planting late is a Nash equilibrium.
c. The best response to the other planting early is to plant early.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
1.12
Improve Stick to
English Swahili
Aisha
Ben
Improve
Swahili
Stick to
English
2
0
●
0
4
1
1
●
4
2
Disagreement Games: Conflict about how to coordinate
We use the Language Game described in Figure 1.15 as an example of a
Disagreement Game. A Disagreement Game illustrates coordination games
in which there are two Pareto-efficient Nash equilibria (which are therefore
Pareto-incomparable), and the players are in conflict over which Nash equilibrium each prefers. So, the players’ problem is to manage to coordinate on
one of the Nash equilibria, or alternate systematically between them, to ensure that no coordination failure results and they do not end up at an outcome
neither would prefer.
Consider two players, a home-language Swahili-speaker (Aisha) and a homelanguage English-speaker (Ben) who have recently met. Each person can
speak the other language, but prefers to speak their home language. They
share many common interests but do not communicate as well as they would
like. Each has two strategies: Stick to your home language or Improve the
other language.
Among the possible outcomes are that he could learn better Swahili and they
could routinely converse in that language; and she could learn better English
and they could converse in English. They do not need to both be fluent in
both languages. So for Aisha, if Ben becomes fluent in Swahili, then her best
response is not to take the time and trouble to improve her English. For Ben,
similarly, if Aisha were to become fluent in English, then there would be little
point in taking the Swahili courses.
The result is two Nash equilibria (Stick to Swahili, Improve Swahili) with payoffs (4, 2) and (Improve English, Stick to English) with payoffs (2, 4). The two
Nash equilibria are both Pareto-efficient because there are no alternative
outcomes which are Pareto-superior to the Nash equilibria.
But, as shown in the payoff matrix 1.15, Aisha would prefer the (Stick to
Swahili, Improve Swahili) Nash equilibrium and Ben would prefer the (Improve
English, Stick to English) Nash equilibrium.
& ECONOMIC INSTITUTIONS
53
Figure 1.15: A Disagreement Game: The players
need to coordinate on an equilibrium, but each
prefers one equilibrium to the other, so there is a
conflict of interest. If they fail to coordinate on one
of the Nash equilibria because of the conflict of
interest, the outcome will be a coordination failure.
54
MICROECONOMICS
- DRAFT
The Disagreement Game is similar to the Assurance Game in that:
• There are two Nash equilibria
• Both players do better if they coordinate (that is, speak the same language
at one or the other of these equilbria)
The Disagreement game differs from the Assurance Game because:
• Each player in the Disagreement Game prefers one of the Nash equilibria
while the second player prefers the other, while both prefer the Paretosuperior Nash equilibrium in the Assurance Game, so as a result
• the players in the Disagreement Game have a conflict of interest concerning which equilibrium gets selected.
Of course the English speaker would prefer to communicate in her home
language and the Swahili-speaker feels the same way, but they would do
much worse if they did not have any common language at all and if they failed
to coordinate.
Disagreement Games highlight how there can be social interactions with
multiple Nash equilibria, each of which is Pareto-efficient, but there may be
no ’middle ground’ to coordinate on and as a result conflict over who gets
to benefit the most is unavoidable. Both players in the Disagreement Game
would both be worse off out of equilibrium than at one of the Nash equilibria in
the game. They have a common interest in coordinating somehow as opposed
to not coordinating; but their interests conflict in how they coordinate.
Checkpoint 1.14: Language Game
Label the outcomes of the Language Game as in Figure ??, plot them using
axes with the players’ payoffs, and determine which outcomes are Nash equilibria and which are Pareto-efficient.
1.13
Why history (sometimes) matters
As we have seen from Disagreement Games and Assurance Games, strategic
complementarities in games may give rise to more than one Nash equilibrium.
When this is the case we cannot say which Nash equilibrium is our prediction
of how the game will be played. The best the Nash equilibrium concept could
do is to say that the outcome of the game is likely to be one of the (perhaps
many) Nash equilibria.
We need more information to make a prediction. Think about the Palanpur
game, and imagine that all you know is the payoff matrix (not how the farmers
played the game in recent years). Though you would be on solid ground
H I S TO RY One of the first game theoretic
studies of coordination problems – by
David Lewis – was concerned with how we
coordinate on a common language.8
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
& ECONOMIC INSTITUTIONS
55
predicting that it is likely that you’d see both farmers planting either early or
late, you would not have much confidence in which it would be.
But now suppose you were told that last harvest they planted late. Then,
unless they had discovered some way to coordinate a switch to planting early,
you would be correct when you predicted that they would both be planting late
this year too.
When history matters in this sense, we say that outcomes may be pathdependent. When the outcome of a game is path-dependent, without
PATH DEPENDENCE A process is path dependent if the most likely state of something this
period – the fraction of farmers planting early
or late in the example – depends on its state
in recent periods.
knowing the recent history of a social interaction we cannot predict which
equilibrium will occur. So, quite different outcomes – poverty or affluence, for
example – are possible for two interacting groups of participants with identical
preferences, technologies, and resources but with different histories. This is
how "history matters."
The Palanpur payoff matrix describes a poverty trap. A poverty trap occurs
when identical people in identical settings may experience either an adequate
living standard or poverty, depending only on chance events of their histories,
for example were their parents rich or poor, or were they citizens of Norway or
Nigeria. The possibility of poverty traps alerts us to the fact that people may
be rich or poor not because of anything distinctive about their skills, hard work
or other personal attributes, but because of the situation they find themselves
in. Poverty may be inherited as it is in Palanpur not by anything that parents
pass on to children but instead by the inheritance of a common history.
The same is true about other aspects of how we interact in society, for example in the ways our lives may be highly segregated in interacting with people
who differ in the groups with which they are identified, whether that be ethnicity, or religion , or even loyalty in sports teams.
Checkpoint 1.15: Drain the meadow: Name that game
Write down a payoff matrix for Hume’s "drain the meadow" game, with the two
actions open to farmers Adams and Brown being "drain" and "do not drain", and
assuming that the value of the drained meadow (to each farmer) is 5, the value
of the undrained meadow is 3, and if the two farmers jointly work on the draining
it costs them 1 each, while if a single farmer does the draining alone it costs him
3. What kind of game is this? Explain how it might be solved if there were just
two farmers, and why with many farmers (as Hume wrote) it would be " difficult
and indeed impossible" for them to agree on a common course of action and
avoid in a coordination failure.
P OVERTY T RAP A poverty trap occurs when
identical people in identical settings may experience either an adequate living standard
or poverty, depending only on chance events
of their histories. Poverty in this case is a
result of a person’s circumstances.
56
MICROECONOMICS
11
- DRAFT
12
1
10
2
9
3
8
4
7
6
5
(a) The circle as a clock
1.14
(b) The baseline
Application: Segregation as a Nash Equilibrium among people who prefer integration
Segregated communities – whether on grounds of ethnicity, race, religion,
or class – are often the basis of inter-group prejudice and hostility and the
systematic denial of equal dignity to all citizens. Segregation often results from
deliberate policies of discrimination by governments, lenders, and citizens,
as illustrated by the apartheid system of enforced racial separation in South
Africa that persisted until 1994 and legally mandated housing segregation in
(c) The citizens’ ideal integrated neighborhood
Figure 1.16: The neighborhood and the citizen’s ideal integrated outcome Panel a is the
"geography" of the neighborhood, showing that, for
example, the citizen at position 2 on the circle has
two immediate neighbors, the people at positions 1
and 3. Panel b shows that the person at position 2
is a Blue and her two immediate neighbors are both
Greens. is just a starting point at which the neighborhood is as integrated as possible in the sense
that the two immediate neighbors of each citizen
are of the other type. Panel c shows the distribution
of types across locations that the citizens prefer:
each citizen has one immediate neighbor of each
type.
the U.S. – the so-called racial covenants the were finally outlawed in 1968.9
But segregation can also result from the uncoordinated decisions of people
who would actually prefer to live in integrated communities.
This counter-intuitive result illustrates the use of the Nash equilibrium concept,
underlining the lesson already learned from the interaction among the Palanpur farmers: There may be more than one Nash equilibrium – one Pareto
superior to the other – and a society can find it it difficult to escape the inferior
equilibrium. The example of segregation is also a reminder – like the case
of the over-fishing Nash equilibria– that the fact that an outcome is a Nash
F AC T C H E C K In Seattle, Washington, what
are termed “racially restrictive covenants”
covering more than 20,000 homes prohibit
sale or rental to particular groups. One
stipulated that, “No person or persons of
Asiatic, African or Negro blood, lineage, or
extraction shall be permitted to occupy a
portion of said property.”10 Racially restrictive
covenants have been illegal since 1968 in
the U.S. and are unenforceable.
equilibrium does not mean that it is something that the players would choose,
if they could coordinate and decide jointly on the outcome.
The set-up of the model
Here is a model. There are two types of people, Greens and Blues, and they
live in homes arrayed around a circle representing a neighborhood. The
homes are identical except that they may differ in the types of the immediate neighbors. The neighborhood is the circle as a whole. A household’s
immediate neighborhood is made up of the two households on either side of
it.
Figure 1.17: Thomas Schelling (1921-2016) was
an American economist who won the Nobel Prize
in economics in 2005 for his contribution to our
understanding of conflict and cooperation. The
model we propose here is based on his work.11
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
A
B
C
(a) Household B has one neighbor of
each type, their ideal situation, is not
dis-satisfied, and will play Do Nothing
A
B
C
(b) Household B has two neighbors of
their own type, is not dis-satisfied, and
will play Do Nothing
If a citizen would like to live at some other location around the circle, they can
switch with some other person currently occupying that position, as long as
the other person is willing. The homes just change occupants with no money
changing hands. We would like to know what the neighborhood will look like
when all the switching that people can do has been carried out, so that the
neighborhood’s composition stops changing. The distribution of types among
the houses around the circle when no citizen can benefit by moving is a Nash
equilibrium.
Greens and Blues have identical preferences about the type of their two
immediate neighbors only. All people in the neighborhood would prefer to
have one neighbor of each type, as is shown in Figure 1.18. But they are
“satisfied” as long as they either have an immediate neighbor of each type
or if both are of their own type. People are “dissatisfied” if both immediate
neighbors are of the other type. An ideal neighborhood, then is shown above
in Figure 1.16 c: Each person has one neighbor of each type.
People have two strategies: “Do Nothing” or “Signal Dissatisfaction.” Signalling
dissatisfaction means being willing to switch positions with another person
– anywhere in the neighborhood – who has also “signalled dissatisfaction.”
People are willing to switch only if they prefer the new location to their old
location. For this reason people of either type will never switch with a person
of the same type. This is because, for example, if a Green is dissatisfied with
her current location and would like to move, all other Greens would be equally
dissatisfied were they to take her position, so no other Green would agree to
a switch. So all of the switches will be with different types: a green will switch
with a blue, but a blue will never switch with a blue and a green will never
switch with a green. This means that switches will change two things:
• the switcher’s new immediate neighborhood: those on either side now
experience having a neighbor of a new type given the switcher’s arrival and
the previously dissatisfied person’s departure
• the switcher’s old immediate neighborhood: those who were previously on
either side of the switcher have a neighbor of a new type given the arrival of
the person with whom the previously dissatisfied person switched.
& ECONOMIC INSTITUTIONS
A
B
57
C
(c) Household B has two neighbors of
the other type, is dissatisfied, and will
play Signal Dissatisfied
Figure 1.18: The preferences of a household
depending on the kinds of neighbors that
surrounds it. Household B will either be satisfied
or dissatisfied depending on the the types of
neighbors they have. B’s choice to play Signal
Dissatisfaction or Do Nothing therefore depends on
the composition of their immediate neighborhood.
58
MICROECONOMICS
- DRAFT
A segregated Nash equilibrium
We begin with 6 Greens and 6 Blues occupying alternating positions in the 12
“houses” at the locations on the circle that are numbered as if from time on
a clock (so, 12 is the top). The twelve homes on the circle are "the neighborhood." We call the assignment of different types to the the homes around the
circle in Figures 1.16 b and c: an allocation. An allocation in this game is an
assignment of homes to the types at a given stage of the game. The allocation
before the game starts is the initial allocation. The allocation after the game
ends is the final allocation.
The game proceeds as follows. At each step, each of the 12 people plays either Do Nothing, or Signal Dissatisfaction. Their choice of a strategy is known
to all other players. Then, one of the twelve citizens is randomly selected and
given the opportunity to make a switch if she can find another person willing
who has also signalled dissatisfaction and is willing to make a switch.
At step 1, for example, we might ask the Green at position 10 o’clock if she
would like to switch. She would, because both of her neighbors are Blues.
Whether she is able to make a switch depends on whether there are others
who have chosen the strategy Signal Dissatisfaction. Because everyone else
is also dissatisfied, she has many choices.
Suppose she switches with her friend and immediate neighbor, the Blue at
position 11, shown by the colors of position 10 and 11 changing from Panel
a Start to Panel b Step 1. The two people are still friends and neighbors, but
each now also has the same type of neighbors on the other side.
Suppose, next, that it is the Blue at position 7 who is picked to stay or switch.
If he plays "Signal Dissatisfaction," he could switch with his friend and immediate neighbor at position 6. We continue this process until either no one
is dissatisfied, or if someone is dissatisfied, there are no others playing the
strategy Signal Dissatisfaction with whom a switch is possible.
This process could continue as shown in the figure, resulting at the end of
5 steps in the completely segregated neighborhood shown in Figure 1.19.
Notice that over the course of the game, a particular home may change hands
more than once. The home at position 7 for example started off occupied by
a Blue who switched with a Green in Step 2, who then switched with a Blue in
Step 4.
At step 6 (not shown), each of the 12 would choose the strategy Do Nothing,
because 8 of them have same type as neighbors only and the other four have
one neighbor of each type. So no one is dissatisfied. As a result there we
observe no further moves: the allocation is stationary (meaning unchanging).
It is a Nash equilibrium.
We could expand the strategies available to the players to allow those with
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
& ECONOMIC INSTITUTIONS
(a) Start
(b) Step 1
(c) Step 2
(d) Step 3
(e) Step 4
(f) Step 5: Complete Segregation
both neighbors of their own type to signal dissatisfaction, and if a willing other
citizen could be found, they could switch with this person so as to have one
neighbor of each type. This will not disrupt the completely segregated neighborhood as long as people prefer to have both neighbors of their own type to
having both neighbors of the other type.
Avoiding outcomes that nobody prefers
The conclusion is not that complete segregation will necessarily be the result.
This is true for two reasons.
• There is also a Nash equilibrium that is integrated rather than segregated.
In Figure 1.16 c, the allocation has each person’s immediate neighborhood
composed of both types. You can confirm that, like the completely segregated allocation, this integrated allocation is also a Nash equilibrium: every
citizen has their ideal immediate neighborhood, so no citizen is dissatisfied
and each are best responding with Do Nothing. This allocation could have
59
Figure 1.19: From integration to a segregated
Nash equilibrium. The figure shows one of many
possible progressions from an integrated non
equilibrium situation to an entirely segregated Nash
equilibrium. Panel a shows the starting point from
the previous figure. In step 1 the Green at position
10 and the Blue at position 11 switch positions,
shown by the double heded arrow, and resulting in
the neighborhood shown in Panel b. The remaining
panels show the successive steps to the final fully
segregated Nash equilibrium.
60
MICROECONOMICS
- DRAFT
come about by the same rules of the game that resulted in complete segregation. This is an example of implementing a desirable allocation within
given set of rules of the game
• The citizens could play the game cooperatively rather than non-cooperatively.
If the citizens had realized that playing the game non-cooperatively could
lead them to a complete segregation outcome that nobody wanted, they
could have acted cooperatively – that is jointly agreed – to implement their
ideal allocation. This is an example of implementing a desirable allocation by changing the rules of the game: agreeing to act jointly was not an
available strategy in the non-cooperative variant of the game above.
The outcome in the segregation model shares three features with a game
from what would appear to be a very different situation: Planting in Palanpur.
• A Pareto inferior Nash equilibrium: There is a Nash equilibrium – planting
late and a segregated community – in which everyone is worse off than
they could be at some other allocation.
• A path-dependent outcome: History matters because an outcome that is
preferred by all participants is also a Nash equilibrium, so if the preferred
outcome were to occur, it could persist.
• A change in the rules of the game can avoid the inferior outcome: By
coordinating their actions – changing the interaction to a cooperative game
– they could escape the Pareto inferior outcome
In these three respects the two interactions – when to plant and where to live
– are not unique or even unusual in these three respects.
Checkpoint 1.16: Segregation as a Nash equilibrium
a. Show that the segregated neighborhood in Figure 3 is a Nash equilibrium.
b. Show that the ‘ideal neighborhood’ in Figure 1 is also a Nash equilibrium.
c. Show that in Figure 1.19 if Step 3 had been different the equilibrium allocation could have been the citizens’ ideal integrated allocation. Which Step 3
switch would have accomplished this result?
d. Suppose that the game was changed slightly so that a dissatisfied person
knows only the “dissatisfaction status” of her immediate neighbors. Show
that, starting with the alternating types status quo (Panel a Start, in Figure
1.19 ) that the neighborhood would then evolve to the ideal distribution.
e. Suppose that in the fully segregated neighborhood case citizens decided to
have a binding referendum to implement the ideal neighborhood (requiring
whatever moves are necessary to bring that about), but because the question of where you live is a sensitive one, they adopt the rule that unanimous
approval of the referendum is required for it to be implemented. Would it be
implemented?
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
1.15
How institutions can address coordination problems
Game theory has given us a catalogue of coordination problems: Prisoners’
Dilemmas, Invisible Hand Games, Assurance Games, and Disagreement
Games (there are many more!). Knowing how the structure of these games
differ will help to diagnose the nature of a coordination problem and to devise
policies and constitutions – changes in the rules of the game – to avoid a
coordination failure.
This is an example of using the concept of equilibrium to understand how to
change an undesirable outcome. The idea is simple: a change in the rules of
the game can eliminate an undesirable Nash equilibrium, so that it is no longer
our prediction of how the game will be played. Instead it may be possible to
make some preferable strategy profile a Nash equilibrium which then could be
the outcome of the game.
A common approach to averting coordination failures is in a Prisoners’
Dilemma to devise policies or institutions that transform the payoff matrix
so that the game is no longer Prisoners’ Dilemma. There are two possibilities
to consider:
• Change the Prisoners’ Dilemma to an Assurance Game
• Change the Prisoners’ Dilemma to an Invisible Hand Game
Changing the Prisoners’ Dilemma game into an Assurance Game means
making the mutual cooperate outcome a Nash equilibrium (it was not in the
Prisoners’ Dilemma) even if mutual defect remains a Nash equilibrium.
The second options is a more ambitious policy objective: converting a social
interaction from a Prisoners’ Dilemma to an Invisible Hand Game. To see
how this might work, remember that the coordination failure that results in the
Prisoners’ Dilemma is a consequence of the fact that in that game players can
take actions that inflict costs on others – negative external effects – that are
not part of their thinking when they decide what to do.
To see that internalizing these external effects can address the coordination
failure, we examine the implementation of a liability rule in the Fishermen’s
Dilemma. Tort is a branch of law dealing with damages inflicted by one person
on another (or another’s property). Among other things, tort law establishes
the responsibility – called liability – of the person inflicting the damages to
compensate the harmed individual. The requirement to compensate the
harmed individual internalizes the external effect.
How would a liability system work in the Fishermen’s Dilemma? Look again at
Figure 1.12, the general form of the Prisoners’ Dilemma. Suppose Alfredo and
Bob decided to jointly adopt “Cooperate” (fish 10 hours) as an agreement. In
their agreement, they also choose to adopt a liability rule requiring compen-
& ECONOMIC INSTITUTIONS
61
62
MICROECONOMICS
- DRAFT
Bob
3
2
12 Hours
3
2
10 Hours
2
2
(a) Liability rule with numbers
(Defect) (Cooperate)
3
3
10 Hours
(Cooperate)
12 Hours
(Defect)
Alfredo
10 Hours
(Defect) (Cooperate)
12 Hours
Alfredo
10 Hours
(Cooperate)
Bob
12 Hours
(Defect)
y − (w − x)
w
●
●
w
w
w
y − (w − x)
z
z
(b) General liability
sation be paid to the other party if one’s actions result in lower payoffs than
would have occurred had the agreement to cooperate (fish only 10 hours)
been observed.
With the liability rule the following will happen:
• If Alfredo defects on Bob (plays Fish 12 hours), Alfredo initially gets y as
before
Figure 1.20: Fishermen’s Dilemma with a liability
rule. Players can implement a desired outcome
by Transforming Property Rights using a liability
rule (the harm a player does to another player is
deducted from their payoff). This payoff matrix is
based on Figure 1.12 modified by the liability rule.
Alfredo’s payoffs are listed first in the bottom-left
corners and shaded blue. Alfredo’s best responses
are shown by the solid point. Bob’s are listed
second in the top-right corners and shaded red; his
best responses are shown by the hollow circle.
• But then Alfredo must compensate Bob for the costs his defection has
inflicted, that is, Alfredo must pay compensation sufficient to give Bob a
payoff of w (which is the payoff Bob would have obtained had Alfredo not
violated the agreement)
• If both players defect, they both gain z
• But, then they must compensate the other player by a transfer of w
z.
We can use these changes to the payoffs to construct a transformed payoff
matrix. The transformed payoff matrix for Alfredo’s and Bob’s payoffs is given
by the entries in Figure 1.20.
Did the improved property rights succeed? Because y
w + x < w by the def-
inition of a Prisoners’ Dilemma, Cooperate is now a best response to Cooperate and (Cooperate, Cooperate) is a Nash equilibrium. Cooperate is also a
best response to Defect (because w > z), so Cooperate is the dominant strategy with the liability rule in force, and (Cooperate, Cooperate) is the dominant
strategy equilibrium. The Prisoners’ Dilemma game has become an Invisible
Hand game through the adoption of a new set of property rights.
Redefining property rights – to take account of liability for damages – can
implement a Pareto-efficient outcome by inducing each player to account for
R E M I N D E R For a Prisoners’ dilemma we
need y > w > z > x and x + y < 2w. It’s the
second one that’s important here.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
how his actions effect the other player. By redefining property rights to include
the liability of the damages (external effects) that one inflicts on others, we
have transformed the game to an Invisible Hand game.
M-Note 1.1: The mathematics of the liability rule
Refer to Figure 1.20. For the original game to be a Prisoners’ Dilemma requires:
• y > w > z > x,
• x + y < 2w,
• and has the Nash equilibrium (Defect, Defect) with payoffs (z, z).
The following inequalities must help us to think through the logic of the liability rule:
• With the transformed payoffs the Nash equilibrium must be (Cooperate, Cooperate)
with payoffs (w, w).
• The condition for x + y < 2w arises because we require that for w > y (w x). We
can re-arrange the condition because w > y
w + x, therefore, by adding w to both
sides we get 2w > y + x as we said we need to assume.
Be sure that you can identify the intermediate step here that we have skipped in the matrix: For the combination (12 Hours, 10 Hours) and (10 Hours, 12 Hours) the payoff to
the player playing 10 hours is w because it was x + (w
with x, but were then rewarded with the transfer for (w
the agreement. x + (w
x) (that is, the players started
x) by the other player breaking
x) = w! Notice too, that both of the players receiving z in the
(12 Hours, 12 Hours) occurs because they both compensate the other with (w
z
(w
z) + (w
z) and
z ) = z.
Checkpoint 1.17: Limited Liability by the Numbers
Use the model of the liability rule in Figure 1.20 to complete the following tasks.
a. Re-draw the payoff table, but substitute in the values for x, y, w and z from
Figure 1.10. Hint: The payoffs should only be 2s and 3s.
b. Solve your new game using best response analysis (the circles and dots
method) to find the Nash equilibrium of the game. What is it? Explain.
c. Does either player have a strictly dominant strategy? Is there a dominant
strategy equilibrium? Explain.
1.16
Game theory and Nash equilibrium: Importance and caveats
We have started off this introduction to modern microeconomics with game
theory. The reasons are that
• Many important economic relationships – in labor markets, families, credit
and financial markets, between citizens and governments, among neighbors, between nations seeking to address climate change, and many more
– are strategic, and require the tools of game theory.
& ECONOMIC INSTITUTIONS
63
64
MICROECONOMICS
- DRAFT
• Focusing on people as actors often in conflict with each other, but also
sharing common interests, is essential to economics as a social science,
and game theory allows us to do this.
• How these interactions work out depends on the institutions that regulate
them, and game theory allows us (even requires us) to be very specific
about the varieties of possible rules of the game under which we now
interact, and how we might change these rules for the better.
For game theory the Nash equilibrium is a key economic idea and it provides
a way to answer the question: what will be the outcome if each of the actors
adopts a strategy that will not lead any other actors to change what they
do?
In many situations the Nash equilibrium among players who seek to maximize
their own material payoffs provides a good prediction of what we observe in
the real world. But not always.
• Extreme individualism: Overlooked opportunities to coordinate: If the
two fishermen were fishing 12 hours each and the details of the situation
were such that they could just agree to fish 10 hours instead, then we
would be mistaken to use the Nash equilibrium of this game to predict
what the players would do. In this case the rules of the game have been
inadequately described: there was a third strategy that was overlooked,
namely agree with the other to fish 10 hours as long as the other agreed
to the same. With the game modified in this manner, the Nash equilibrium
would provide a good prediction: both Bob and Alfredo would choose
"Agree" as long as they believed that the other would do the same and that
the agreement would be enforced.
• Self-interest: Overlooked payoffs not in the form of one’s own material
gains: Even if for some reason they could not agree on a common course
of action, if Alfredo and Bob were brothers and each cared about how
much fish the other had to feed his family, then the two would not define
their payoffs in the game simply as amount of fish they each caught individually, but would include the consequences of their actions on the outcome
for the other fisherman. So we would have to rethink the payoff matrix. In
this case it is the payoff matrix that was mistaken. It no longer describes
what the players are trying to maximize, that is the incentives that are
shaping their behavior.
• Selection among many equilibria: As we have seen in the Planting in
Palanpur Game, there may be more than one Nash equilibrium, so the
prediction that the result of the interaction will be a Nash equilibrium is
insufficient. We need to know more about the situation – such as the recent
history of the people involved – to make a prediction.
E X A M P L E When people can bind themselves to common agreements or when they
have values other than maximizing their own
material gains, we need to modify the game
as we will see later in this chapter, and in
Chapter 2.
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
• Dynamics: Sometimes we are more interested in the process of change
itself than in the stationary end point of this process. Whether studying how
the people of Palanpur might make a transition from poverty to affluence,
or how the polar sea ice might collapse due to climate change we are often
interested in how the economy works when it is not in equilibrium.
1.17
Application: Cooperation and conflict in practice
If all that is needed to address a coordination failure is to require that people
pay the costs that their actions impose on others then why are coordination
failures so common?
Over-exploitation of fisheries is an international problem that humans as
a world community have failed to solve. Many over-exploited fisheries will
not recover for a long time. But local communities and groups of fishermen
have found ways to combat over-fishing and we can learn many lessons from
what they do. Many groups – from farmers to fishermen – face equivalent
problems worldwide. These outcomes provide a concrete motivation to study
the Prisoners’ Dilemma Game and other coordination problems.
What we learn from these games is that an effective liability rule requires two
things:
• The injured party or the courts have to have verifiable information (that is
information sufficient to enforce the liability aspect of the property right) and
• There has to be a court or some other body willing to and capable of enforcing the contract.
When we turn from game theory to the study of real fishing communities
we find that both conditions are unlikely to be met, which is why the overexploitation of fisheries continues in many cases.
• Limited information. The lack of verifiable information is common in social
interactions and this limits the policies that governments or private actors can design in response to the persistence of Pareto-inefficient Nash
equilibria.
• Conflicts of interest. Governments may not have the capacity or the will
to enforce the necessary policies especially in cases where doing this
would impose costs on a powerful group. An example is the failure of many
countries to address the problem of climate change, which is in part the
result of the fossil fuel companies’ opposition to putting a sufficiently high
price on carbon emissions.
Fishing communities, of course, are not acting out a tragic script, as were
the herders in Hardin’s tale about the tragedy of the commons. They are not
& ECONOMIC INSTITUTIONS
65
66
MICROECONOMICS
- DRAFT
prisoners of the dilemma they face. Real fishermen are resourceful and seek
solutions to the problem of over fishing.
• Lobstermen in the U.S. state of Maine limit how many lobster they catch
using highly local restrictions on who can set traps where (the state government provides the legal framework for this).12
• Turkish fishermen allocate fishing spots by lottery and then rotate the spots
so the distribution is fair.13
• The fishing community of Kayar in Senegal adopted the rule that only one
trip to the fishing grounds per day is permitted (a bit like Alfredo and Bob
limiting their hours of fishing) and appointed a committee to check that the
rule was being observed. They also limited the number of boxes of fish that
could be offloaded by a single canoe.14
• Shrimp fishermen in Toyama Bay, Japan have a rule that they offload their
daily catch at the same time and place, so that the size of each boat’s
catch would no longer be asymmetric information.15
These rules and practices based on small local fishing communities are often disrupted by the entry of other groups who are not bound by the local
rules. Conflicts of interest within the local community also sometimes limit
the effectiveness of attempts to limit the catch. One reason is that restrictions
on fishing are often supported as a way to raise the wholesale price of fish
and hence the incomes of fishing families. But fish sellers – who buy the fish
wholesale at the port and then sell to local consumers – would profit if they
could pay less.
The rules regulating access to fishing that we see in existence are a small
selection from a much larger set of rules that people have tried out at some
point. What we see are the institutions that have succeeded well enough to
allow the communities using them to persist and not to abandon their rules.
The persistence of such rules does not require the rule to implement a Paretoefficient outcome, it only requires that the rule be reproduced over time by
people adhering to it. By this reasoning, even if the rules of the game do not
implement Pareto efficient outcomes, we might expect a fishing community
that has hit on a way of sustaining cooperation in the long run to do better
in competition with groups that over-fish, and that successful groups may be
copied by other groups.
Checkpoint 1.18: Institutions and Palanpur
Supposing that the only voters involved in approving the Palanpur village council’s decision to require planting early were themselves farmers, explain why
they would unanimously support the measure. What would happen if after implementing the law requiring early planting one year, the next year the law was
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
taken away?
1.18
Conclusion
The classical institutional challenge which we stated originally was “How can
social interactions be structured so that people are free to choose their own
actions while avoiding outcomes that are worse for everyone than they could
be?” With the terms you have learned this can now be re-phrased “How can
social interactions be structured so to avoid Pareto-inefficient Nash equilibria
resulting from people’s free choice of their own actions?” The Fishermen’s
Dilemma is an example of a challenging coordination problem because an
inefficient outcome is the unique Nash equilibrium.
To study a game and its likely outcomes and also how to improve these outcomes we have proceeded in three steps:
• First, use the Nash equilibrium concept to identify one or more likely outcomes of the game
• Second, use Pareto comparisons to identify outcomes that are "worse for
everyone," and
• Third, devise changes to the relevant institutions – the rules of the game –
or that would shift the population to a superior Nash equilibrium either preexisting (as in the case of Planting in Palanpur, or the segregation case) or
novel (as with the transformed Prisoners Dilemma Game).
We have illustrated the third step by a legal remedy, the introduction of tort
liabilities for damages in the Prisoners’ Dilemma Game so as to internalize
the external effects accounting for the coordination failure. But an adequate
analysis of coordination problems and their possible mitigation must do two
things that we have not yet taken up:
• Evaluate outcomes according to their fairness and other ethical criteria and
• illuminate the far broader range of how markets, governments, private
associations like firms or neighborhoods and other bodies might jointly
accomplish this task. We take up this challenge in Chapters 2, 4 and 5.
Making connections
Institutions and the rules of the game: To predict or explain the outcome
of a social interaction, it is essential to know the “rules of the game” that
determine who knows what and when, who gets to do what and when and
as a result who gets what.
& ECONOMIC INSTITUTIONS
67
68
MICROECONOMICS
- DRAFT
Equilibrium: Equilibrium describes an outcome that will persist until some
aspect of the situation changes as a result of externally caused changes. A
Nash equilibrium is a special kind of equilibrium widely used in economics.
External Effects: People often take actions without considering the effects
of these actions on others. The resulting external effects – positive and
negative – pervade social interactions.
Pareto inefficiency of Nash equilibria: A common result is that the outcomes of social interactions (the Nash equilibria of the games) are Paretoinefficient meaning that opportunities for mutual gains remain unrealized.
Rents: When players interact they face opportunities for mutual benefit, or
common interest. But this creates opportunity for rents and for a conflict
over how the benefits resulting from the interaction will be distributed.
Policy and changes in the rules of the game: Improving property rights (such
as making people legally responsible for the harms to inflict on others) can
lead people to internalize external effects. Other institutions that would
facilitate people making decisions to act jointly can also provide solutions
to coordination problems. Policy may result in a shift to a Pareto-efficient
equilibrium.
Positive feedbacks, increasing returns, and strategic complementarity: Often
players strategies are strategic complements due to positive feedback
and increasing returns. As a result, in some social interactions there may
may be many equilibria as in the Assurance Game and the Disagreement
Game.
Important Ideas
institutional challenge
fallback
Pareto-superior/inferior
coordination problem
next best alternative
Pareto efficient
player
best response (weak/strong)
(economic) rent
strategy
dominance (weak/strict)
payoff
Dominant strategy equilibrium
institution
Nash equilibrium
interdependence
positive external effect
negative external effect
Prisoners’ Dilemma
(non-)cooperative games
optimization
Tragedy of the commons
Invisible Hand Game
Assurance Game
Disagreement Game
Increasing returns
Strategic complement/substitute
Path dependence
Poverty trap
Liability rule
positive feedback
S O C I E T Y : C O O R D I N AT I O N P R O B L E M S
Mathematical Notation
Notation
Definition
p
a player’s payoff
Note on superscripts: A, B: individuals.
Discussion questions
See supplementary materials.
Problems
See supplementary materials.
& ECONOMIC INSTITUTIONS
69
2
People: Self-interest and Social Preferences
DOING ECONOMICS
How selfish soever man may be supposed, there are evidently some principles
in his nature, which interest him in the fortune of others, and render their happiness necessary to him, though he derives nothing from it, except the pleasure of
seeing it.
Adam Smith, The Theory of Moral Sentiments (1759) Chapter 1.
Chicago, a city famous for its pizza, sports, jazz, and its skyline, built its fortune on the farming of the state of Illinois. Today Illinois farmers use high tech
machinery and advanced business plans, some cultivating a thousand acres
of land or more. But many of the farmers don’t own the land they cultivate;
they rent land and work it as a sharecropper. Sharecroppers are farmers –
"tenants" – who pay the owners of land a share of the total harvest that they
cultivate.
In the mid-1990’s, over half of the contracts between farmers and owners
were sharecropping agreements, and in Northern Illinois 95 percent of these
contracts stipulated a fifty-fifty division of the crop between the owner and the
sharecropper. An equal split of the crop means that a tenant on fertile land will
have higher income for the same amount of effort and other inputs.1 Because
a tenant on fertile land will reap a larger harvest than a tenant on poor quality
land, the fifty-fifty sharecropping contract presents us with a puzzle.
Here’s the puzzle: if half of the crop on poor quality land is sufficient to attract
tenants, why should the owners of good quality land give up half of the crop to
their tenants? Those tenants must be earning more than what was necessary
to get them work the owner’s land. So, why don’t the owners of the better land
propose a tenant’s share less than half, giving the tenants just enough so that
This chapter will enable you to:
• Understand that people make decisions
based the actions open to them (constraints), which of these possible actions
they believe they must take (beliefs) to
bring about the outcomes they most
prefer (preferences).
• Use this approach to analyse economic
situations involving risky outcomes,
bargaining, and contributing to the public
good.
• Analyse sequential games and games
with multiple Nash equilibria, showing
how being the first mover in these games
may confer advantages on a player.
• Explain how the experimental method
along with this "preferences, beliefs, and
constraints approach" is used to study
economic behavior.
• See how changes in the rules of the
game can result in better outcomes for
all.
• Describe the experimental and other
evidence that people’s preferences
go beyond self-interest and include
generosity, reciprocity, fairness and
hostility toward others .
• Understand that these other-regarding
preferences are as much part of what we
consider to be rational action as is self
interest.
• Give examples of the importance of
social norms and culture for decisionmaking and economic policy-making.
they are willing to farm the land?
We would expect owners to insist on lower tenant’s shares to sharecroppers
on higher quality land and offer higher shares to sharecroppers on low quality
Figure 2.1: Farming in Illinois is big business.
72
MICROECONOMICS
- DRAFT
land. Because land varies in quality by small gradations, this would result
in a pattern of sharecropping contracts with a range of shares depending
the land quality. But they is not what we see. Almost all of the contracts are
fifty-fifty.
Illinois sharecropping contracts allow the sharecroppers on good land to
receive income attributable to the superior land quality, income the owners
would otherwise have received if the owners had insisted on a lower tenant’
s share on the high quality land. The fifty-fifty split effectively transfers millions of dollars annually from owners to sharecroppers simply because of the
fifty-fifty division. This is not a peculiarity of Illinois. Fifty-fifty is the norm in
sharecropping around the world.
Rice cultivation in West Bengal, India during the 1970s provides another
example. There, poor illiterate farmers in villages isolated by impassable
roads for much of the year and lacking electronic communication eked out a
bare living on plots that average just two acres rather than the thousand-acre
plots farmed in Illinois. The Indian farmers shared one similarity with farmers
in Illinois: the division between sharecroppers and owners was fifty-fifty in over
two-thirds of the contracts.2
Why was the contract the same in these distant parts of the world? The short
answer is that where most contracts are fifty-fifty, that particular division is a
social norm, something people feel they are morally obliged to respect. The
fact that around the world land owners respect a social norm that overrides
their material self-interest tells us that many people are committed to acting
fairly, being treated fairly and conforming to ethical standards of appropriate
behavior.
But the sharecroppers in Illinois and West Bengal, like farmers everywhere,
are also trying to make a decent living, or even to become affluent. They are
not simply following social norms. They carefully weigh alternative methods of
cultivating their crops at the least possible cost and marketing their harvest at
the highest possible price.
2.1 Preferences, beliefs and constraints
Understanding economic behavior requires a model that takes account of
what people care about ( for example, the farmers’ incomes, and also their
desire to uphold social norms) and how from the actions they are able to
undertake, they adopt those that are they think will bring about desired results.
We will develop a model of economic behavior based on:
• constraints: the feasible set of actions, meaning actions that are open to
us,
H I S TO RY In 1848 the British philosophereconomist John Stuart Mill noted the striking
global pattern of equal division in crop
sharing, calling it “the custom of the country”
and “the universal rule.”3
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
Beliefs
Preferences
(which actions result
(evaluation of
in which feasible outcomes)
feasible outcomes)
73
Figure 2.2: Preferences, beliefs and constraints.
The actor may choose from a set of feasible actions
(the constraint set on the left). Combining that set
with her beliefs about the outcome produced by
each of the actions in the the constraint set, she
then has a set of outcomes that she believes are
feasible, depending on her choice of an action.
From all of these outcomes in the set believed to
be feasible, she identifies the one that is ranked
highest according to her preferences and then
takes the action that she believes will bring about
this outcome.
Set of
Constraints
(Feasible set
of actions)
outcomes
believed to
Choice of
an action
be feasible
• beliefs: our understanding of the outcomes that will result from the actions
that are open to us,and
• preferences: our evaluation of the outcomes that we believe will result
from the actions we take.
This is called the preferences, beliefs and constraints approach.
The relationship between these three elements of the preferences, beliefs,
and constraints approach is described below and is shown in Figure 2.2.
Game theory, which you have already studied, is an important example of the
preferences, beliefs and constraints approach.
Constraints: limits on action
From a long list of things she might consider doing, constraints define a more
limited possible set of actions, namely the shorter list of all of those so called
feasible actions she can carry out. In the game theory introduced in the previous chapter the constraint was the set of possible actions, that is, a list such
as "Fish 10 hours", "Fish 12 hours" or "Plant early", "Plant late".
Constraints may be imposed by personal limitations, by laws of nature, or
by the force of law. A constraint can also reflect a decision by the actor to
eliminate some action from the feasible set of actions on moral grounds,
irrespective of the payoffs. Examples are keeping promises, or not committing
murder.
In Table 2.1 we give examples of how the preferences beliefs and constraints
approach can be applied.
The list of feasible actions set by constraints need not be just a list of particular actions, like drive or take the bus. When marketing their output (first row of
E X A M P L E The preferences, beliefs and
constraints approach is sometimes called
rational choice theory or the rational actor
model, but we prefer the more specific label
that we use here as it identifies the the three
important elements making it up.
P REFERENCES , BELIEFS , AND CONSTRAINTS
APPROACH According to this approach, from
the feasible set (which includes all of actions
open to the person given by the economic,
physical or other constraints she faces) a
person chooses the action that she most
values as given by preferences, in light of
given her beliefs about the actions that will
bring about the outcome.
74
MICROECONOMICS
- DRAFT
Constraints
Actor
(feasible set
of actions)
Beliefs (information
Feasible Outcomes
about which actions
(states that could
will result in the
result from
preferred state)
the actions)
The demand curve
Firm owner
High or low prices
(how quantity
depends on price)
Urban resident
Ordering a meal
Drive or
How many others
take the bus
will drive
The menu;
your budget
Simple: just order
the best you
can pay for
Various levels
of profits
Travel time
Meal quality,
money left over
the table), the owners of a firm, for example, can set any price they like (anywhere from 0.00 by penny increments up to some very high number).
Wealth, the availability of credit, and the prices of goods impose constraints
Preferences
(ranking of
all outcomes)
Maximize profits
Minimize travel time
Maximize utility
Table 2.1: Applications of the preferences,
beliefs, and constraints framework. Real choice
situations are typically not as simple as Figure 2.2.
The urban resident, for example, may care both
about travel time to work and his carbon footprint.
on an actor’s consumption. The institution of private property also imposes
limits: it means that theft is not an option for increasing your consumption.
Given private property and in the absence of gifts or other transfers from a
government, the total amount of goods and services you can consume is limited by wealth and how much you can borrow. So when we study someone’s
consumption, their budget constraint is a critical factor as people have a certain budget determined by wealth, access to credit, and prices all limiting how
much they can buy.
Beliefs: Translating actions into outcomes
Beliefs are a person’s understanding of the outcomes that her actions will
bring about.
In many cases what I must do to get the outcome that I prefer depends on
what other people do. I would like to spend the evening with friends, but where
I should go to make it happen depends on where I think my friends will go.
Given that I cannot communicate with my friends (the batteries to their phones
have run out), my action (where I will go) will therefore depend on my belief
about where I will find my friends.
In Table 2.1 the owners of firms are not constrained to set any particular price,
but if they want to translate their choice of a price into what they care about –
profits – they must form an opinion about the number of units they will be able
to sell at each price. This is the demand curve, and it expresses the owners’
beliefs about the relationship between their action (the price) and an outcome
(how many goods they will sell).
In the game theoretic approach of Chapter 1 beliefs were expressed in the
B ELIEFS A persons understanding of the
relationship between her actions and the
outcomes that will occur as a result of her
actions are her beliefs. Beliefs are thus a
causal mapping from the actions one can
take to the outcomes that will occur. Where
the outcomes of actions are not known with
certainty, beliefs include probabilities of
results occurring.
E X A M P L E The word "belief" is often used to
refer to spiritual matters ("religious beliefs");
but in game theory a belief is a statement
about how the world works, namely what
action is required to bring about some
particular outcome.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
75
solution concept, that is a description of how the game would be played. The
Nash equilibrium as a solution concept, for example, is based on the idea
that players best respond to the play of the other players. This is the basis
of players’ beliefs about how their choice of an action will translate into an
outcome of the game.
Preferences: Reasons for preferring one outcome over another
Preferences are evaluations of outcomes that provide motives for actions. A
person’s preferences are the reason why she takes the action that that she
believes will bring about the outcome that is better than or at least as good
as the others. In Chapter 1, preferences were represented by the payoffs in
games that people played. For each player, a strategy profile was associated
with a number – her payoff – and players chose actions that they believed
would result in the strategy profile with their most preferred (highest payoff)
outcome.
In many games preferences are represented by money payoffs. But more
E X A M P L E While most widely used in
economics, the preferences, beliefs, and
constraints approach is also used in political
science, for example, to understand the
strategies followed by elected officials
seeking to maximize their chances of reelection, in law to design criminal or civil
penalties to effectively deter illegal activity,
and even in biology to study the evolution
of genes, modeled as if they are "trying to"
increase their numbers.
P REFERENCES Preferences are evaluations
of outcomes that provide motives for taking
one course of action over another.
broadly, preferences represent the favorable (positive) or unfavorable (negative) feelings a person has about an outcome that lead them to try to make
an outcome happen (high payoff) or that lead them to try to avoid an outcome
(low payoff). Preferences include:
• tastes (food likes and dislikes, for example),
• habits (or even addictions),
• emotions (such as anger and disgust) often associated with visceral reactions (such as nausea or an elevated heart rate),
• social norms (for example, those that induce people to prefer to be honest
or fair), and
• psychological tendencies (for aggression, extroversion, and the like).
The difference between preferences and beliefs is simple. A preference says:
I like the outcome X more than the outcome Y. A belief says: I believe I can
get X to happen if I do some action Q.
Self-regarding and other-regarding preferences
A feature of the preferences, beliefs, and constraints approach is that it allows
us to model choices based on the entire range of preferences whether they be
entirely self-regarding, caring for others (wishing them well or wishing to harm
them), or reflecting religious commitments.
A key distinction about our preferences is whether in evaluating the results
that we believe our actions will bring about (the right hand part of Figure
2.2) we think about the results that we ourselves experience only, or do we
OTHER - REGARDING PREFERENCES A person
with other-regarding preferences when
evaluating the outcomes of her actions takes
into account the effects of her actions on the
outcomes experienced by others as well as
the outcomes she will experience.
76
MICROECONOMICS
- DRAFT
also consider the results that are experienced by others. This gives us two
categories of preferences:
• If we think only about the results experienced by ourselves, we have selfregarding preferences
• If we also think about the results experienced by others, then we have
other-regarding preferences.
Is the the same thing as "selfish" and "unselfish" preferences? No.
Abraham Lincoln is said to have remarked: “When I do good, I feel good.
When I do bad, I feel bad. That is my religion.” Does this mean that Lincoln’s
"good" acts were in fact self-regarding because they made him feel "good?"
SELF - REGARDING PREFERENCES When
choosing an action, a self-regarding actor
considers only the effect of her actions on
the outcomes experienced by the actor, not
outcomes experienced by others. A self
regarding actor ignores the external effects
of her actions on others.
That does not follow. He had other-regarding preferences leading him to act
differently than if he cared only about the outcomes that he personally experienced. In the preferences, beliefs and constraints model all actions are
motivated by preferences, so doing a preferred thing cannot be termed "selfish" without making all behavior selfish by definition. That is why we use the
term self-regarding rather than "self-interested" or "selfish." For example, if
you (like Lincoln) enjoy helping others, and you act on these preferences,
does this mean you are selfish (because, for example that’s what gives you
a sense of leading a good life). No, it does not. You are acting on your preferences, but they are other-regarding because you enjoy trying to make the
results that others experience be what they would want. Of course otherregarding preferences include feelings of altruism towards others, but they
also include negative feelings about others, such as envy, spite, racism and
homophobia.
H I S TO RY In 1977 Amartya Sen wrote "Rational fools" in which he pointed out that the
preferences beliefs and constraint approach
ignores the importance of promises, what
he called commitments. The reason is that
the approach seeks to explain behavior entirely on the basis of the actor’s anticipation
of what her actions will bring about in the
future. Honoring a past commitment – not
because she would otherwise feel guilty in
the future, but because it is the right thing to
do – cannot be modeled in the preferences
beliefs and constraints approach.
In sections 2.8, 2.9 and 2.12 we provide some evidence from experiments on
the kinds of other-regarding preferences and how common they are across
the world.
"Rationality"
The term rationality in economics means acting on the basis of:
• Complete preferences This means, that for any pair of possible outcomes
that a person’s actions may bring about, A and B, it is the case that the person prefers A to B or B to A or is indifferent between the two. Preferences
are not complete if there is some other pair, say A and D for which none
of the above three comparisons can be made: to the three statements “I
prefer A to D," “I prefer D to A" and “I am indifferent between A and D" the
person responds "none of the above."
• Consistent preferences If an individual with consistent (also called transitive) preferences prefers a bundle of goods A to another bundle B, and
bundle B to a third bundle, C, they also prefer A to C.
R ATIONAL A rational person has complete
and consistent (transitive) preferences and
can therefore rank all of the outcomes that
their actions may bring about (better, worse,
equal) in a consistent fashion.
C OMPLETE PREFERENCES Complete
preferences specify for any pair of possible
outcomes that a person’s actions may bring
about, A and B, whether A is preferred to B,
B is preferred to A or they are equivalent.
Using the symbolic notation for preference:
A B, or B A, or A ⇠ B.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
A person with complete preferences, which requires only that she can rank all
pairs of outcomes, might nonetheless violate the consistency assumption. So
she could prefer A to B, B to C, and C to A. All that matters for completeness
is that she can rank each pair.
77
INDIFFERENCE When a person is indifferent
between two outcomes, it is because those
outcomes provide them the same payoffs,
or the same expected payoffs. As a result,
a person will not care which of the two (or
more) outcomes they obtain between (or
among) which they are indifferent.
In the heading at the start of this section, we put quotation marks around rationality to underline the difference between how economists use the term and
how it is generally used, that is to mean "based on reason." In everyday usage, the term "rational" often means something like the intelligent and perhaps
even amoral pursuit of one’s own interests. But in economics, as you can see
from the above definition, it means something entirely different.
• Rationality does not say anything about what it is that the person values: A
completely generous and ethical person is rational as long as her preferences are consistent and complete.
• Rationality does not mean being intelligent or well informed: The beliefs
that (along with preferences) determine the choices a person makes need
not be true.
Moreover, people with incomplete preferences would hardly be called "irrational" in the ordinary meaning of that term, meaning "not logical" or "unreasonable." Ask yourself if your preferences are complete for the following
outcomes: express preference or indifference over which of your two dearest
friends will be tortured to death. If you were to say "I cannot rank those two
outcomes, nor am I indifferent between them" you would not be "rational" by
the economic definition, but nobody would think your behavior was bizarre
either. We might be more inclined to worry about the person who would be
able to make such a ranking.
Checkpoint 2.1: Why beliefs matter
Considering the coordination problems studied in Chapter 1
a. Explain why in the Assurance Game representing planting in Palanpur why
the action a farmer takes to bring about the preferred outcome depends on
the farmer’s belief about what other farmers will do.
b. In the same game explain why the farmer who believes most other farmers
will plant late, will also plant late.
c. Explain why Ben’s belief about what Aisha will do matters for how he will play
in the Disagreement Game.
d. Are there any games you have learned so far in which beliefs about what the
other does did not affect the outcome of the game?
C ONSISTENT ( OR TRANSITIVE ) PREF ERENCES Preferences are consistent
(transitive) if whenever an individual prefers
a bundle of goods A to another bundle B,
and bundle B to a third bundle, C, they also
prefer A to C. Using the symbol A B to
mean "A is preferred to B" and the symbol )
to mean "implies", the condition for consistency can be written as: A B and B C
) A C.
78
MICROECONOMICS
- DRAFT
2.2 Taking risks: Payoffs and probabilities
Beliefs become especially important in cases where we have to take some action without knowing for sure what the outcome will be. You make many of this
kind of choices every day, from the important choices of what to study at university, to more trivial choices like whether to take an umbrella to class. The
theory of decision-making in these cases rests on the idea that the evaluation
of how good a course of action is depends on
• how much the decision-maker values each of the possible but uncertain
outcomes of the action and
• the decision-maker’s beliefs about how likely each is.
Here we introduce a basic concept for decision-making with risk – expected
payoffs – that will be used throughout the book. In Chapter 13 we return to
the topic of risk including preferences about risk taking per se and the value of
insurance.
The value of uncertain outcomes: Expected payoffs
There are two possible but uncertain outcomes of the action "take an umbrella
to class," namely, "keep dry walking home in the rain" and "carry the umbrella
to and from class without even opening it, because it does not rain." The feasible actions of the decision maker are just: take the umbrella or not.
According to the preferences beliefs and constraints approach, the decision
maker assigns numbers indicating how much she values each of the possible
four outcomes shown in Table 2.2. These numbers give the ranking of the
four possible outcomes: [Don’t take the umbrella, No rain] is better than [Take
the umbrella, Rain] and so on. But if they are to provide a framework for
making a decision when you do not know for sure if it is going to rain or not,
Figure 2.3: Amartya Sen (1933- ). Image Credit:
National Institutes of Health, Public Domain.
the numbers have to be more than a ranking. They have to indicate how much
the actor values each of the possible four outcomes. So for example taking the
umbrella when it rains is 5 times better than not taking the umbrella when it
rains.
We call these numbers the payoffs to each of the four possible outcomes.
The likelihood of uncertain outcomes: Beliefs
Only one of these two uncertain events will occur. Whether at the end of
the day, it turned out to have been a good idea to have brought the umbrella
said to be contingent on (meaning: depends on) whether it rains or not. The
C ONTINGENCY The payoff to the outcome
of a decision is said to be contingent if
something affecting the payoff may or may
not happen. The payoff in this case is said to
depend on a contingency.
payoff to the two actions in this case is said to depend on a contingency. The
contingency in this case is whether or not it rains and the payoff to taking the
umbrella is contingent on (depends on) its occurrence.
P ROBABILITY DISTRIBUTION A probability
distribution for n contingent outcomes of a
decision is a list of non-negative numbers
{P1 , P2 , . . . , Pn } that add up to 1. These
probabilities express the decision-maker’s
degree of belief about the likelihood that
each of the of n contingent outcomes will
occur.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
Uncertain event
(contingency)
Action
Rain
No Rain
Take the umbrella
15
8
Don’t take the umbrella
3
20
79
Table 2.2: Two contingencies (rain or don’t
rain) and two actions (Take the umbrella, or
Don’t). The payoffs correspond to the coincidence
of an action and a contingency, so Anoushka
receives 15 if she plays Take the umbrella when the
contingency is Rain, and she receives 8 if she plays
Take the umbrella and the contingency is No rain.
When you decide what to study at university before knowing what kind of work
you’ll do after, you’re making choices about contingencies too: do you go risky
and study in drama, or do you go safe and take in accounting? In this case,
the contingencies include the the uncertainty about how good you will be at
the field you choose and your chance of getting a job in your field.
The theory of decision-making about risky outcomes assumes that a decisionmaker, call her Anoushka, has beliefs about the probabilities (Pi ) that each
of the contingencies i = 1, . . . , n will occur. Her beliefs can be based on
observation, on empirical studies, guessing, experience, or superstition. They
R ISK AND UNCERTAINTYThe term risk
is conventionally used in economics to
describe situations where the probabilities of
the possible outcomes are known. The term
uncertainty describes situations where the
decision-maker does not know and cannot
learn these probabilities.
need not be correct.
For simplicity we assume contingencies with just two outcomes (like "it rains"
or "it does not rain" above). The basic principles of decision-making are the
same no matter how many contingent outcomes there are. In this case, we
use the symbol P for the probability the contingency occurs, understanding
that 1
P is the probability the contingency does not occur.
The decision rule: Maximize expected payoffs
Often we must take an action prior to the realization of the outcome – you do
not know with certainty what will happen, that depend on an uncertain events
– called a contingency – that may or may not happen. But you have to make a
choice anyway. To take account of the "action now, contingency later" aspect
of the decision problem we distinguish between:
• Expected payoff : how much the actor values taking the action given her
beliefs about the probability that the contingency will occur and
• Realized payoff : how much she values the possible outcomes that may
happen, that is, after the contingency has been realized ("realized" here
means really happened, or actually occurring).
The expected payoff of an action is the basis for her choosing one course of
action over another: Anoushka chooses the action with the highest expected
payoff. Here is how she can calculate expected payoffs.
For each contingency, i, and each action she can take, x, Anoushka knows the
payoff of taking action x conditional on i happening, which we write as p (x|i).
For example, if i is the contingency of rain in the afternoon, and x is the action
of taking her umbrella with her in the morning, then her realized payoff is
p (x|i) associated with her having the umbrella when it rains. The vertical line
H I S TO RY In 1947 John von Neumann and
Oskar Morgenstern showed that how much
we value some action that we can take,
when the outcome of the action is subject to
some risky contingency can be expressed
as a weighted sum of how much we value
the alternative outcomes of our actions that
depend on the realization of the contingency,
the weights being the probability of each
outcome occurring if we take the action.
80
MICROECONOMICS
- DRAFT
| is read "conditional on", or "given", so p (umbrella|rain) is Anoushka’s payoff
to having the umbrella conditional on, or given, rain in the afternoon.
For a contingency with two outcomes – numbered 1 and 2 – we have to
consider only two payoffs and the corresponding probabilities of each,
((p (x|1), P), (p (x|2), 1
P)). For example, if Anoushka’s payoffs for the
four possible outcomes of her actions are as in Table 2.2, and the probability
of rain in the afternoon as 0.6, her system of contingent payoffs for taking the
umbrella is (15, 0.6), (8, 0.4)). These numbers can be interpreted as follows:
since there is a 60% chance of rain, Anoushka has a 60% chance of receiving
a payoff of 15 if she takes the umbrella. Further, this means that there is a
40% chance of no rain, therefore, if Anoushka takes the umbrella she has a
40% chance of having a 8 payoff.
The expected payoff to an action x given a system of contingent payoffs
is the weighted average of the payoffs for each contingency where the
weights are the the probability of each contingency being realized if the action x is taken. We abbreviate the expected payoff to choosing x given the
probabilities (Px ) of contingencies 1 and 2 being realized as E (px , Px ) =
E ((p (x|1), P), (p (x|2), 1
Expected Payoff
P))
E (px , Px )
= Pp (x|1) + (1
P)p (x|2)
(2.1)
Equation 2.1 expresses the fact that the greater the probability of an outcome,
the greater its weight in the weighted average calculated by the expected
payoff. For example, Anoushka’s expected payoff to taking the umbrella,
assuming the payoffs and probabilities above, would be 0.6 · (15) + 0.4 · (8) =
9 + 3.2 = 12.2.
Calculating expected payoffs with probabilities is essential to understanding
strategic interactions, such as the games we introduced in Chapter 1. But in
games – that is strategic interactions with other people – the contingencies
are the strategies chosen by the other player, not something like whether it
rains.
Checkpoint 2.2: Basis of probability assessments
a. Imagine that you are rolling two six-sided dice with sides corresponding to
one of each of the numbers 1, 2, 3, 4, 5, and 6. You calculate the sum each
time you roll the two dice simultaneously, for example, 1 + 2 = 3. Explain why
the probability of getting a total of 7 from rolling the two dice is 1/6.
b. What is the expected payoff if you get paid $5 for rolling a sum of 6 or 8 on a
roll of the two dice and $0 otherwise?
c. Go back to Table 2.2, what would Anoushka’s expected payoff to not taking
the umbrella be given the probability of rain being P = 0.6?
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
Figure 2.4: Planting in Palanpur: An Assurance
Game. Aram’s payoffs are listed in the blue bottomleft corner. Bina’s payoffs are listed in the pink
top-right corner. Aram’s best response to Bina’s
choice of strategy is indicated by a black dot in
the relevant cell, while Bina’s best responses are
indicated by hollow circles. Using the dot and
circle method you can confirm that the the Nash
equilibria of the game (any cells with both a dot and
a circle) are (Plant Early, Plant Early) and (Plant
Late, Plant Late), with payoffs (4, 4) and (2, 2). The
Plant Early Nash equilibrium is Pareto-efficient.
The Plant Late equilibrium is not.
Bina
Early
Late
Aram
Early
●
Late
3
4
0
4
0
3
●
81
2
2
2.3 Expected payoffs and the persistence of poverty
In games like the Prisoners’ Dilemma which have a dominant strategy equilibrium, the action that will maximize your payoffs does not depend on what the
R E M I N D E R A dominant strategy equilibrium
is a strategy profile in which all players play a
dominant strategy.
other player does, so it does not matter that you do not know what the other
will do.
But if – like most games – there is not a dominant strategy equilibrium, then
your best response depends on what the others do, and we need to take
account of this in our decision making rule. We can use expected payoffs
to understand the choice of which strategy to play in an Assurance Game,
like a farmer’s choice between Planting Early or Planting Late in the Planting
in Palanpur game. The game is shown in Figure 2.4 to remind you of the
game’s structure. The payoffs in each cell indicate how much the farmer
values outcome resulting from the strategy profile given by the particular row
and column.
As you know, the game has two Nash equilibria: (Early, Early) and (Late,
Late). Recall that (Early, Early) is Pareto-superior to (Late, Late).
The Plant Early equilibrium is also the payoff-dominant equilibrium.
An
equilibrium is payoff dominant when no other equilibrium exists that is Paretosuperior to it. The Pareto-efficient Nash equilibrium in an assurance game is
payoff-dominant. In our example, Plant Early is payoff-dominant because the
payoffs in this equilibrium exceed the payoffs for both players in the Plant Late
equilibrium.
As we observed in Chapter 1, Palanpur farmers plant late even though a
Pareto-superior alternative exists. To see why this occurs, think of what the
other player will do as a contingency. We can then say that the degree of
belief that other farmers will plant early can be expressed as a probability,
P.
A farmer believing with probability P that the other farmer will plant early and
probability (1
P) that the farmer will plant late is an example of decision-
PAYOFF - DOMINANT E QUILIBRIUM An equilibrium is payoff dominant when no other
equilibrium exists that is Pareto-superior to it.
The Pareto-efficient Nash equilibrium in an
assurance game is payoff-dominant.
82
MICROECONOMICS
- DRAFT
Bina
Late
Aram
Early
Early
(P)
Late
(1−P)
4P
0(1−P)
3P
2(1−P)
Figure 2.5: Aram’s view of Planting in Palanpur.
The figure shows Aram’s payoffs only (and not the
payoffs to any other farmers) and his belief about
the probability they will occur. Aram’s payoffs are
multiplied by the probability of that cell occurring
based on the probability that Bina plays that
strategy. The other farmers play Plant Early with
probability P and Plant Late with probability 1 P.
We can calculate Aram’s expected payoffs to
each of his strategies by adding the payoffs to a
given strategy against each of the other farmers’
strategies. Plant Early: p̂Early = 4 · P + 0 · (1 P)
Plant Late: p̂Late = 3 · P + 2 · (1 P).
making under risk, since the farmer assigns probabilities to a contingency, in
this case, the other farmer’s behavior. We do not explore where these beliefs
about probabilities come from, but we can imagine that he farmer will form
beliefs based on what other farmers tell him or on the basis of their behavior
in past planting seasons. We will include just Aram and Bina in the game, but
remember we use only two players to simplify our analysis of what is really a
much larger population of many people like Aram and Bina.
If Aram believes that the probability of Bina planting early is P we can con-
ˆ,
M - C H E C K We label expected payoffs as pi
which is sometimes written E [p ].
struct his expected payoffs to each of his strategies, each part of which is
shown by Figure 2.6.
We use a "hat" on a variable to mean ’expected,’ so p̂ reads "p hat." Using
these probabilities, Aram’s expected payoff (E (p )orp̂ ) to playing Plant Early
is:
p̂ = p̂ (plant early))
= Pp (plant early|others plant early)
+(1
P)p (plant early|others plant late)
Aram’s expected payoff to planting late is:
p̂ (plant late)
= Pp (plant late|others plant early)
+(1
P)p (plant late|others plant late)
An expected payoff-maximizing farmer will choose to plant early or late depending on which expected payoff is higher. As Figure 2.6 shows, for Aram,
which action this will be depends on the probability that he thinks Bina will
plant early. The vertical axis is the expected payoff to each strategy: Plant
Early or Plant Late. The horizontal axis is the probability, P, that Bina plants
early: from left to right P goes from P = 0 (the Bina plants late with certainty)
to P = 1 (the Bina plants early with certainty).
The two upward-sloping lines plot the expected payoffs to the two strategies,
Plant Early and Plant Late, for Aram the farmer making a decision and how
I NDIFFERENCE P ROBABILITY In a two-by-two
game, let P be the probability that player A
attributes to B playing one strategy and 1 P
the probability A attributes to B playing the
other strategy. Then Pi is the value of P such
that such that player A’s expected payoffs
to to playing each of her two two strategies
are equal. In this case player A is therefore
indifferent between playing the two strategies
(which is why we use the letter i subscript.).
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
^
Expected payoff, π
Strategy with higher
expected payoff at P = 1 2
is risk−dominant
4
Equal expected
payoffs at Pi = 2 3
3
2.67
2.5
r
2
i
Expected payoff to
^L
Plant Late, π
Expected payoff to
^E
Plant Early, π
P = 1 2 Pi = 2 3
0
1
Probability Bina will play early, P
they depend on the probability that he believes Bina will plant early (that is, for
each value of P).
The blue line graphs the equation for the payoff to the strategy Plant Early
which is p̂Early (P) = P · 4 + (1
P) · 0 = 4P. When the probability the other
farmer plants early is zero, i.e. P = 0, the payoff to Plant Early is zero. When
the probability the other farmer will Plant Early is 1, i.e. P = 1, the payoff to
Plant Early is 4.
We can draw the expected payoff line for Plant Late in the same way, where
the expected payoff to Plant Late is p̂Late (P) = P · 3 + (1
P) · 2 = 2 + P
depicted in green, and where pLate (P = 0) = 2 and pLate (P = 1) = 3. We can
then interpret the expected payoffs as follows:
• Plant Late provides a higher expected payoff for all P < 2/3.
• Plant Early provides a higher expected payoff when P > 2/3.
• The expected payoffs to the strategies are equal at the indifference probability Pi = 23 (where a farmer is indifferent between Plant Early and Plant
Late).
The result is that Aram will choose Plant Late as long as he believes that the
probability that Bina will Plant Early is less than two-thirds. Bina, facing the
identical situation, has the same decision rule: Plant Late unless you think that
Aram is going to Plant Early with a probability of at least two-thirds.
They will remain poor even though, had they somehow started of both planting
83
Figure 2.6: Aram’s expected payoffs to planting
early or late depend on his belief about the
probability that Bina will plant early. Aram
evaluated the expected payoffs to his strategies
based on the probability that Bina will play
Early. The indifference probability where the
two strategies have the same expected payoff is
Pi = 2/3, and the payoff to Planting Late is greater
than the payoff to Planting Early for P = 1/2. The
intercepts of the vertical axes are the payoffs in the
payoff matrix for the planting game in Chapter 1
(Table 2.4).
84
MICROECONOMICS
- DRAFT
early, they would have been much better off. The poverty trap in which they
find themselves is not the result of rudimentary technology or infertile soil.
What they lack is the "social technology" that would allow them to coordinate
on the Pareto-superior strategy profile, planting early. Their poverty is due to
the rules of the game, which make coordination difficult.
In this example we have assumed that both Aram and Bina had some idea
(maybe a guess) of the likelihood that the other would plant early. They faced
risk (they had some information on the probability of the contingent event),
but not uncertainty (no information at all). Decision making under uncertainty
is especially important in the field of climate change, where there are some
contingencies for which there is no way to assign probabilities of their occurrence.
M-Note 2.1: Expected Payoffs for Planting in Palanpur
To understand the expected payoffs and the indifference probability in the game, we need
to answer the following questions:
a. What are the expected payoffs to planting early and planting late (using payoff Table
2.4) for a farmer in Palanpur who believes the probability of Bina planting early is P,
with 0 < P < 1?
b. What value of P leads to an equal expected payoff to planting early and late?
We can work out these answers using the following steps:
• If Aram plants early and Bina plants early, Aram’s payoff is 4 (with probability P)
• If Aram plants early and Bina plants late, Aram’s payoff is 0 (with probability 1
P)
• If Aram plants late and Bina plants early, Aram’s payoff is 3 (with probability P)
• If Aram plants late and Bina plants late, Aram’s payoff is 2 (with probability 1
P)
The expected payoff to planting early is the weighted average of the two contingencies
(Bina plays Plant Early or Plant Late) with the weights equal to the probability (P) of Bina
playing Plant Early and (1 P) of Bina playing Plant Late:
• Expected payoff for planting early p̂Early (P) = 4 · P + (0) · (1
• Expected payoff for planting late: p̂Late (P) = 3 · P + 2 · (1
P) = 4P.
P ) = 2 + P.
• To find the indifference probability at which expected payoffs are equal: p̂Early (P)
p̂Late (P)
=
• Which is the condition: P = Pi = 23 , the indifference probability.
2.4 Decision-making under uncertainty: Risk-dominance
What is the farmer facing uncertainty to do? Economics does not have a very
good answer.
A two-person risk-dominant equilibrium
Economists often use what is called the "principle of insufficient reason"
when a player has no information on which to place a probability on some
H I S TO RY The "principle of insufficient
reason" due to the Swiss mathematician
Jakob Bernoulli (1655-1705) states that if we
have no information on which to estimate the
probability that one of two contingencies will
occur, we should consider them to be equally
likely. Not everyone finds this satisfactory.
John Maynard Keynes found it "paradoxical
and even contradictory."4
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
85
contingency. This principle holds that the farmer who has no information on
likely strategy choice of his neighbor will assign equal probability to the two
events and hence use the probability P = 12 that the other will plant early.
What is termed the risk dominant strategy is that which yields the highest
expected payoff when a player attributes equal probability to the two actions of
the other player.
Using this definition in the Planting in Palanpur game, a farmer who assigns
the probability P = 12 to the outcome that the other farmer will Plant Early will
himself Plant Late because his expected payoffs are 2 = 12 · 4 to Plant Early,
and 2.5 = 12 · 3 + 12 · 2 to Plant Late.
Thus Plant Late is the risk-dominant strategy, that is, the strategy that maximizes the farmer’s expected payoffs when P = 1/2. You can confirm this
by going back to Figure 2.6: at P = 1/2 the green line (expected payoff to
planting late) is above the blue line (expected payoff to planting early). Because this is true for the other farmer as well, both farmers Planting Late is the
risk-dominant equilibrium.
Planting late in the Planting in Palanpur game is risk dominant because planting early when the other plants late is much worse (you get zero rather than
the payoff of two you would have received had you also planted late) than
planting late when the other plants early (you get three rather than the four
you would have received had you also planted early).
Checkpoint 2.3: Risk dominance and the worst case outcome
a. Redraw the expected payoff line for planting early with the payoff to planting
early when the other plants late to be even worse than shown in the figure,
e.g. -2 instead of 0.
b. In this case what is the indifference probability?
c. What is the least payoff to planting early when the other plants late, that
would make planting late no longer risk dominant?
A risk dominant equilibrium in a large population
Instead of thinking about only two farmers, we can interpret as portraying a
population of farmers in a village like Palanpur itself, who all face the same set
of incentives for planting early and late. Like Aram and Bina, all the farmers
face a coordination problem: doing well if they all plant early and doing poorly
if they all plant late. How well each farmer does depends on what the others
do, so if a minority of farmers plants early while the majority plants late, then
those who planted early while others planted late will have their seeds eaten
while the others will get an adequate harvest. The farmers are therefore
involved in a many-player coordination problem.
We can re-purpose Figure 2.6 such that the horizontal axis is the fraction
R ISK DOMINANT STRATEGY The strategy that
yields the highest expected payoff when the
player attributes equal probability to the two
actions of the other player.
86
MICROECONOMICS
- DRAFT
^
Expected payoff, π
4
3
2.67
2.5
r
2
i
Expected payoff to
^L
Plant Late, π
Figure 2.7: Fraction of farmers planting early. P
is the fraction of farmers playing Plant Early. 1 P
is the fraction of farmers playing Plant Late. In the
case of the population as a whole, the indifference
probability (or in the case of a population of
players, the indifference fraction) shown at point
i with fraction Pi corresponds to the fraction of
the population at which the players are indifferent
between the strategies (Plant Early or Plant Late).
In the case of the whole population, point i is also
the tipping point: when a fraction of the population
less than Pi plays Plant Early all farmers will want
to play Plant Late; when a fraction of the population
greater than Pi plays Plant Early all of the farmers
will want to play Plant Early.
Expected payoff to
^E
Plant Early, π
P = 1 2 Pi = 2 3
0
1
Fraction playing Plant Early, P
of the population going from 0 to 1 who choose Plant Early, P (reading left
to right) as shown in Figure 2.7. Reading the horizontal axis from right to
left, it measures the fraction of the population who Plant Late (1
P). The
payoff lines in the figure have the same interpretation as before: They are the
expected payoffs for any one of the large number of identical farmers in the
village. The probabilities translate to population fractions too:
• P < 23 : When less than two-thirds of the population choose Plant Early (i.e.
more than one-third play Plant Late), the Plant Late strategy will get him
a higher expected payoff. At any fraction P < 23 all farmers will Plant Late
and all farmers will end up with a payoff of 2.
• P > 23 : When more than two-thirds of the population select Plant Early (i.e.
less than one third select Plant Late), the Plant Early strategy has a higher
expected payoff. At any fraction P > 23 all farmers will Plant Early and all
farmers will end up with a payoff of 4.
• P = 23 : At two-thirds Planting Early and one-third Planting Late, the expected payoffs are equal. The point at which the expected payoffs are
equal is a tipping point as a small change will drive all players to adopt
one or the other strategy: Plant Early or Plant Late.
Now imagine that as in the village of Palanpur virtually all of the farmers have
been planting late year after year (maybe even generation after generation).
There would not be much uncertainty about what fraction of the population
would plant late the next planting season. Each of the farmers would hold the
belief that P is close to zero and as a result they all would plant late, confirming their beliefs. The belief that almost nobody would plant early sustains both
the low income of the farmers, and perpetuates the belief itself, which year
after year turns out to be correct.
T IPPING P OINT An intersection of the
expected payoffs to strategies shows
a tipping point when a small change in
population fractions playing a strategy results
in a feedback loop driving the game to one
of the extremes, either P = 0 or P = 1. We
describe it as a tipping point since a small
push either way will "tip" the outcome to one
extreme equilibrium or the other.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
87
In the Fishermen’s Dilemma the best outcome for one of the players is the
worst for the other, so there is a conflict of interest between the two. And this
contributes to the difficulty of finding some way of coordinating so as to avoid
over-exploitation of the fishing stock. This is not the problem in the Assurance
Game. There is no conflict of interest: All of the Palanpur farmers prefer
the outcome when they all Plant Early to any other outcome. Their failure
to implement the mutually desired outcome is the result of their inability to
coordinate on planting early, for example when all are planting late to make a
joint decision to all change to planting early.
What may seem to be a minor tweak to the rules of the game under which the
farmers are interacting can help them escape their poverty trap.
2.5 Sequential games: When order of play matters
When we looked games involving risk and uncertainty, we saw that players could end up selecting risk-dominant strategies that implement Paretoinefficient Nash equilibria. But the game we introduced to model the coordination problem facing Aram and Bina was unlike many real world social
interactions, they were total strangers who had no way of coordinating their
actions, and they acted simultaneously (or at least, without knowledge of what
the other had done.)
But it might be that rather than playing simultaneously, they play sequentially. Playing sequentially is a change in the rules of the game; it represents
a change in the institutions governing their interaction. We will see that this
seemingly small change makes it into an entirely different kind of game possibly even allowing Pareto-efficient outcomes.
To see how this could work, suppose the Planting in Palanpur game the Aassurance Game) is now sequential. Aram moves first (he is called the first
mover) and Bina moves second. How will Aram reason?
He has to think about what Bina will do in response to his planting early or
late. He knows that:
• Bina’s best response to his planting late is to plant late and the best response to his planing early is to plant early, and
• his payoff is greater if they both plant early.
So he will announce that he will plant early, and Bina will respond with planting
early. Rather than being stuck planting late with a small harvest, they have
now solved their coordination failure. How did they manage it?
The answer is that the sequential nature of the game gave them a way of
acting together even if they had no way of actually coordinating. By looking
ahead to what Bina would do later Aram he was able to act so that they would
B ACKWARD INDUCTION Backward induction
is a procedure by which a player in a
sequential game chooses a strategy at
one step of the game by anticipating the
strategies that will be chosen by other
players in subsequent steps in response
to her choice. (Induction here means
causation.)
88
MICROECONOMICS
- DRAFT
Aram
●
Aram
●
Plant
early
Plant
late
Bina ●
Plant
early
Plant
early
●
Plant
late
Plant
early
Bina
Plant
late
Bina ●
Plant
late
Plant
early
●
●
●
●
●
(4, 4)
(0, 3)
(3, 0)
(2, 2)
(4, 4)
(a) Full Game Tree
Aram
●
Plant
early
●
Plant
late
Bina
Plant
early
Plant
late
Bina
Plant
late
Plant
early
Bina
Plant
late
Plant
early
Plant
late
●
(0, 3)
(3, 0)
(2, 2)
(b) Aram’s Reduced Choice of Actions
together implement the single Pareto efficient outcome. What Aram did is
called backward induction, which is is a procedure by which a player in a
sequential game chooses a strategy at one step of the game by anticipating
the strategies that will be chosen by other players in subsequent steps in
response to her choice.
To see how this works when order of play matters (and where backward
induction gets its name), instead of using a payoff matrix (or as we have done
for normal form simultaneous move games) we will use a game tree and refer
to the game as being an extensive form game. Game Trees have the same
basic structure as a normal form games represented by a payoff matrix – they
show the strategy set and the payoffs associated with each strategy profile –
(4, 4)
(1, 3)
(3, 1)
(2, 2)
(c) Fully Solved Game
Figure 2.8: Game tree of the Planting in Palanpur (assurance) game. 2.8 a presents the full
game tree for both players. 2.8 b shows the reduced set of action that Aram considers while using
backward induction to solve the game. 2.8 c shows
the solved game tree with the arrows indicating
the path to the Nash equilibrium (Plant Early,
Plant Early). Aram’s actions are shown by the blue
branches and Bina’s by the red branches. Aram’s
actions are reduced because he has projected
forward in time and used backward induction to
work out what Bina will do: planting early if he
plants early and planting late if he plants late,
therefore reducing Aram’s choices to a payoff of
4 if he plants early and a payoff of 2 if he plants
late. So backward induction leads to the Nash
equilibrium of the game being (Plant Early, Plant
Early) with payoffs (4,4)
except that the tree-like structure tells us something about who moves when;
and a strategy profile is now a path through the branches of the tree.
In the game tree structure, the players move in sequence, with the player
on the top of the tree moving first and the player at the bottom moving last.
A game tree for the sequential version of the Planting in Palanpur game is
shown in Figure 2.8.
Aram is the first player, so he is at the top of the game
tree. Bina is the second player, so she is shown acting after Aram. Each
player’s action – planting early or planting late – is shown alongside a branch
of the tree to indicate which action the player chooses as they move along
that branch. Each player’s payoffs are shown at the end of a branch of the
game tree that indicates a specific path to that end point, Aram planting early,
then Bina planting early; Aram planting early, then Bina planting late; and so
on. The payoffs correspond to those we used in Chapter 1. Because of the
branching tree-like structure of the figure there is only one path to each of the
end points.
In Figure 2.8 on the left-hand side we have the full game tree, showing all the
potential payoffs for the game. Bina is the second-mover and she needs to
decide what to do at each point where she could move. If Aram plants early,
Bina can get a payoff of 4 for planting early, or a payoff of 3 for planting late.
So if Bina is self-interested, then she will plant early when Aram plants early
E XTENSIVE F ORM G AME A game portrayed
by a game tree in which the sequence of
actions by the players is made explicit. The
player at the top of the tree moves first, with
subsequent players moving in sequence
after the first player. Payoffs are shown at
the end of the game tree in player order, e.g.
(Player A’s Payoff, Player B’s Payoff). Refer
to Chapter 1 for the definition of normal form
games.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
(4 > 3).
Bina also has to make a choice between her actions if Aram plants late. Bina
can get a payoff of 0 if she plants early given Aram planting late or a payoff
of 2 if she plants late given that Aram plants late (2 > 0). So if Bina is selfinterested, she will plant late when Aram plants late.
We now know what Bina will do, but what will Aram choose to do knowing
this? Using backward induction and having a belief that Bina is self-interested,
Aram will have a choice between a reduced set of payoffs, shown in the central panel of Figure 2.8: either 4 if he plants early or 2 if he plants late. So if he
is self-interested, then he will choose to plant early. As a result, the only Nash
equilibrium of the game is (Plant Early, Plant Early) with payoffs (4, 4).
Checkpoint 2.4: Back in Palanpur
89
E X A M P L E The timing of a sequential game
does not depend on the exact actions being
taken in that sequence, but can depend on
commitments to those actions being taken.
For example, in the second quarter of the
year a company could commit to the pricing
strategy it will follow in the third quarter and
if they believed the company’s commitment,
then other companies have to respond to
that commitment, even if it’s not the third
quarter yet. Similarly, a professor commits to
a policy in her syllabus even if her students
haven’t written a midterm exam or solved a
problem set yet. The student must respond
to the professor’s committed actions and
work out what he will do in response to her
commitment. To design the syllabus the
professor using backward induction thought
through what a student would do in response
to her commitments in the syllabus.
Making the game sequential solved the problem for Aram and Bina. But would
that work for the couple of hundred families in Palanpur? Suppose some order
of play was determined and that the first family had announced that they would
plant early. Would the second family then follow? And the third? Explain why or
why not?
2.6 First-mover advantage in a sequential game
Being first mover did not give Aram any particular advantage in the Planting
in Palanpur Game, it just allowed him and Bina to coordinate on the Paretoefficient Nash equilibrium. The result would have been the same had Bina
been first mover.
But sometimes it is advantageous for a player to move first; this person then
has what is called first-mover advantage.
Think about the Disagreement Game from Chapter 1. Recall that two players, Aisha and Ben, have a disagreement over which (or perhaps both) of
them should study to improve the language spoken by the other. Both prefer
when they are good at speaking some language. But, Aisha prefers that it
be Swahili and Ben prefers that it be English. What happens in this game
when Aisha is the first mover rather than when they both move simultaneously?
Considering the game tree in Figure 2.9, we can solve the game by backward
induction and see that the Nash equilibrium of the game is (Stick to Swahili
(for Aisha), Improve Swahili (for Ben)) with payoffs (4, 2). The outcome (Improve English, Stick to English) which was one of two Nash equilibria in the
simultaneous version of the game is no longer a solution in the sequential version of the game. Aisha does better as a first-mover because she obtains her
F IRST- MOVER ADVANTAGEA player who can
commit to a strategy in a game before other
players have acted is a first mover. This
limits the outcome of the game to a strategy
profile made up of his chosen strategy and
to the other players’ best responses to it,
which may result in higher payoffs for the first
mover. This is called first-mover advantage.
90
MICROECONOMICS
- DRAFT
Aisha
Aisha
Improve
English
Stick to
Swahili
Ben
Improve
Swahili
(4, 2)
Ben
Stick to
English
(0, 0)
Improve
Swahili
(0, 0)
Improve
English
Stick to
Swahili
Ben
Stick to
English
(2, 4)
Improve
Swahili
(4, 2)
(a) Full Game Tree
preferred outcome. Ben would have benefited in the same way had he been
first mover.
The reason why being first mover gave Aisha an advantage is that the normal
form game has two Nash equilibria – one preferred by Aisha and the other by
Ben. In the sequential game the first mover determines which of the two Nash
equilibria will occur. Once Aisha has moved and has established that she will
Stick to Swahili (and not try to Improve English), Ben needs to choose his
action. Ben needs to take Aisha’s move as given. He must therefore choose
his best response to Aisha choosing Stick to Swahili. Given that he would like
to communicate with Aisha, his best response is to Improve Swahili.
We have not asked why Aisha rather than Ben was first mover in the Disagreement Game and why it was Aram rather than Bina in Planting in Palanpur. We showed only that differences in a person’s position in the game gave
them advantages. In this case Aram and Aisha – the first movers – had strategy sets that gave them the capacity to commit to a strategy in advance.
Aram’s first mover status did not allow him to benefit at Bina’s expense; but
this was not the case with Aisha, her first mover status gave her an advantage
over Ben.
First movers in a modern economy are more like Aisha:
• Employers: they commit to the wage, job requirements and working conditions; workers – actual and prospective – best respond to that.
• Banks and other lenders: they set to the interest rate, repayment schedule
and other aspects of a loan contract. Borrowers and would be borrowers
best respond to that.
• Owners of major companies: in the U.S. Walmart, Amazon, Apple – com-
Ben
Stick to
English
Improve
Swahili
(0, 0)
(0, 0)
Stick to
English
(2, 4)
(b) Solved Game Tree
Figure 2.9: Game tree of the Language (Disagreement) game. The left-hand side presents
the full game tree for both players. The right-hand
side figure shows the solved game tree with the
arrows indicating the path to the Nash equilibrium
(Stick to Swahili, Improve Swahili). Aisha’s actions
are shown by the blue branches and Ben’s by
the red branches. Aisha’s actions are reduced
because she has projected forward in time and
used backward induction to work out what Ben
will do: Stick to Swahili if he plays Improve Swahili
and Improve English if he plays Stick to English,
therefore reducing Aisha’s choices to a payoff of
4 if she plays Stick to Swahili and a payoff of 2 if
she plays Improve English. So backward induction
leads to the Nash equilibrium of the game being
(Stick to Swahili, Improve Swahili) with payoffs
(4,2). The outcome favors Aisha over Ben and
therefore conveys a first-mover advantage.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
91
mit to prices and delivery schedules. Consumers best respond.
The fact that people occupy different positions in our economy – Employers
and workers, lender and borrowers – interacting under rules of the game that
give some first mover status and other special advantages is an important
part of the explanation of inequality of wealth and income as we will see in
Chapters11,12, 13, and 15.
Checkpoint 2.5: Ben has the first-mover advantage
a. Consider the sequential Disagreement Game shown in Figure 2.9. Re-draw
the game tree, but with Ben as the first mover rather than Aisha. Show that
(Improve English, Stick to English is the Nash equilibrium of the game.
b. Assuming that the payoffs in the Disagreement Game are in hundreds of
dollars and that you are Ben, how much would you pay for the privilege of
being first mover a) if otherwise Aisha would be first mover, and b) if the
game were to be played simultaneously (so that there is no first mover)?
2.7 Social preferences: Blame Economic man?
The characters in our economics episodes – Aisha and Ben, Aram and Bina –
care exclusively about their own payoffs. For them a "best response" is simply
a "best-for-me response." Is that why they have had difficulty overcoming the
coordination failures they face? The answer, we will see, is that being concerned about how your actions affect others will help to address coordination
failures, but will not be sufficient.
Homo economicus or "economic man" is the term economists have used
to designate an entirely self-regarding and amoral actor, a person who is
not motivated by either a concern for others, or a desire to conform to any
ethical principles. The term is often put in italics to parallel the biological
terminology for a species (like Homo sapiens). Homo economicus, however is
a fictional character or ideal type representing one possible variety of human
behavior.
Models based on Homo economicus have provided predictions about behavior that are borne out by empirical studies that range from how American
windshield installers and Tunisian sharecroppers respond to different work incentives to the effect of taxes on cigarette consumption. But, as we shall see,
Homo economicus is not an accurate depiction of how people behave:
• People volunteer for fire fighting, delivering food to the sick during a pandemic, and other dangerous but socially beneficial tasks, and contribute
substantial sums to charity.
• People participate in joint activities such as strikes or protests even knowing that their individual participation is unlikely to affect the success of
H I S TO RY The idea of basing economics
on the assumption that people are entirely
self-regarding –"solely as a being who
desires to possess wealth" goes back to the
last of the great classical economists, John
Stuart Mill author of Principles of Political
Economy (1848), considered to be the first
economics textbook in the English language.
He considered this view of people to be "an
arbitrary definition of man."5
92
MICROECONOMICS
- DRAFT
the event and that, if successful, the benefits would be widely shared, not
confined just to those people participating in the protest.
• People donate blood to strangers.
• In public opinion polls and in voting, people support taxes that transfer
incomes to the poor even when they are sufficiently rich and unlikely ever
to benefit directly from these policies.
Motivated by these and similar observations a augmented by controlled experiments about human behavior (that we will review below), economists
have revised our assumptions to recognize that people are capable of ethical,
generous, and other motivations as well as self-regarding motives. This is
important because as you learned in the first chapter, coordination failures
occur because we fail to take adequate account of the effect that our actions
have on others. Our concern for others can help to internalize these external
effects whether it be our willingness to curb our carbon footprint or willingness
to protest for causes whose benefits would be widely shared.
But coordination failures cannot be blamed entirely on people seeking to
maximize their own payoffs. Think again about the real farmers in Palanpur, all
planting late when they could all do better if they all switched to planting early.
Suppose one of those farmers was deeply concerned about the poverty of his
entire village, and wished to improve living standards for everyone. He could
not do this by individually planting early.
Now suppose that every villager shared his concerns for all members of their
community. Each one would know that their own decision to plant early would
change nothing (except that their seeds would be eaten by the birds). What
has captured the people of Palanpur in a poverty trap is not that they care
only about their own harvest (they surely care about others’), but their inability
to come to a common agreement to plant early. Their poverty stems from a
problem of institutions, not motivation.
To understand individual behavior and its social consequences we need an
approach that allows for the full range of human motivation.
Checkpoint 2.6: Homo economicus goes to the polls
a. Given that it costs time to cast a vote (going to the voting station, standing in
line, and the opportunity cost of your time), do you think a person with Homo
economicus preferences would vote in most elections? Why or why not?
b. In what circumstances do you think someone would someone with the
preferences of a Homo economicus vote?
While answering these questions, think about the beliefs the person with Homo
economicus preferences would have about the probability his vote will be important to the outcome of the election.
R E M I N D E R Remember that in Chapter
1 we saw how ’internalizing the external
effects’ means getting people to pay for the
external costs they imposed on others and
this resulted in the fishermen choosing to
cooperate and fish less in the Fishermen’s
Dilemma.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
93
And so, while Homo economicus is among the kinds of actors this approach
considers, there are other characters, representing other sides of human
behavior such as generosity, fairness, reciprocity and spite.
What these four aspects of behavior have in common is that they are otherregarding: the outcomes that the actor considers in choosing an action include
things experienced by others, not just outcomes affecting the actor herself.
Here are some common forms of other-regarding preferences:
• Those with altruistic preferences, such as basic generosity, are motivated
to help others even at a cost to themselves, they place a positive value on
the well-being or payoffs of others.
• Inequality-averse or fairness-based preferences motivate people to seek to
reduce unjust or unfair economic differences even if the actor is herself a
beneficiary of these differences.
• A person with reciprocal preferences is motivated to help others who have
themselves behaved generously or upheld other social norms, and also to
punish those who have treated others badly.
• Spite and ‘us versus them’ distinctions that place a negative value on
outcomes experienced by others, often motivate hostility towards members
of religious, racial, ethnic and other groups. Therefore a negative outcome
another person experiences, can result in a positive value for someone
who feels spiteful.
The term "social preferences" is used to describe all types of other-regarding
preferences.
Checkpoint 2.7: Social Preferences & Social Norms
a. Give an example of a preference you have that is not self-regarding.
b. Can you think of any social norms that lead you to act in an other-regarding
way?
c. Suppose that Aram and Bina (in the Planting in Palanpur Game) were of
different religions between which there is hostility, so that each would gain
some pleasure from the misfortunes of the other. Can you show how this
could change the game so that instead of having the Pareto-efficient mutual early planting as one of its two Nash equilibria, it becomes a prisoners
dilemma with planting late as the dominant strategy equilibrium.
2.8 Experiments on economic behavior
Suppose you wanted to know if someone has altruistic preferences. How
would you find out? Would you ask her? Well, that could provide some information, but merely asking might not be entirely convincing , because many
INEQUALITY AVERSION is a preference for
more equal outcomes and a dislike for both
disadvantageous inequality that occurs when
others have more than the actor and and
(to a lesser extent typically) advantageous
inequality that occurs when the actor has
more than others
94
MICROECONOMICS
- DRAFT
people would like others to think they are altruistic even when they are not, so
they might lie.
What about observing her behavior and comparing her behavior to how others
behave? Such observation might be informative, but if we see people behaving differently that could be because the people we observe have different
beliefs or different constraints, not different preferences.
Economists use experiments to study preferences because at least ideally this
allows us to control for (hold constant) the constraints and beliefs of the individuals to focus on the nature of preferences. Experiments allow economists
to implement the ceteris paribus – all else equal – assumption that we think
is so important when we are trying to identify causes and consequences of
some change or difference.
To understand how common different types of preferences are, and how they
affect our behavior, economists use laboratory experiments in which subjects
C ETERIS PARIBUS is a Latin term that
means "other things equal." When we held
another player’s strategy constant in Chapter 1 to find a player’s best response we
were using the ceteris paribus assumption. Similarly, when we use calculus and
mathematically hold other variables constant we are employing the ceteris paribus
assumption.
play games designed to elicit the nature of their motivations.
Experiments play a central role in science: they allow predictions made from
theories to be tested empirically. This has has been done, for example, with
the prediction that players in a Prisoners’ Dilemma experiment choose the
dominant strategy equilibrium. (You will see what happens in this experiment
below.)
In some sciences, experimenters can control almost all relevant conditions in
their environment, the laboratory. Their subjects can be anything from yeast
cells for a biochemist, to fruit flies for a zoologist.
In economics, however, our subjects are people asked to make decisions or
choices in the experiment. It is much harder to control for the various fac-
F AC T C H E C K Behavioral experiments are a
recent addition to economists’ empirical tool
kits; but they have been used in psychology
for almost a century and a half. The main
innovations that economists have made to
experimental social science are the use of
game theory to clarify the role of beliefs and
preferences and nature of incentives and the
common use of monetary payoffs.
tors that affect human behavior than to control the chemical environment of a
colony of yeast cells. Experimental evidence carries little weight unless the experiment can be replicated, that is, repeated by different researchers reaching
the same results.
We use a specific vocabulary when we talk about behavioral experiments in
economics. The following terms will come up often:
• Subject/participant: A subject or participant is a person who participates in
an experiment.
• Endowment: The endowment is an initial amount of money or tokens
later converted to money that subjects receive at the beginning of the
experiment, and later make decisions about in the experiment.
• Incentives: The fact that players stand to win material rewards in varying
degrees depending on how they play the experimental game means that
the experiment mimics many real economic interactions..
REPLICATION When other researchers
independently repeat an experiment and
reach the same results they have replicated
the experiment. When an experiment
can be replicated we know that its results
are reproducible. Science is founded on
reproducible evidence.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
• One-shot vs. repeated: A one-shot experiment occurs once and subjects
make one decision in the experiment as a whole and are paid for that one
decision. A repeated experiment involves subjects making repeated decisions often with information about the play of others on previous rounds,
sometimes with the same other subjects in a group or sometimes with
different subjects.
Here is an example of the importance of using results from experiments to
test a theory. Subjects with self-regarding preferences are predicted to defect
in a one-shot Prisoners’ Dilemma game because defection is the dominant
strategy. But in Prisoners’ Dilemma experiments, the proportion of players
who cooperate rather than defect is commonly between 40 and 60 percent.6
This means the prediction based on the assumption that people are entirely
self regarding was borne out for some but far from all of the subjects. The
finding therefore provoked some rethinking of the predictions based on the
assumption that people are entirely self-regarding.
Many subjects prefer the mutual cooperation outcome and are willing to take a
chance on the other player also not defecting, rather than the higher material
payoff they can obtain by defecting when the other cooperates. When subjects defect, experimental evidence suggests it is because they dislike being
taken advantage of, not because defection is the payoff maximizing strategy
independently of the other participant’s actions.
2.9 The Ultimatum Game: Reciprocity and retribution
Observing substantial levels of cooperation in the Prisoners’ Dilemma game
was a shock to the standard Homo economicus assumptions. But the experiment that has sparked the perhaps the greatest reconsideration of the Homo
economicus model is the Ultimatum Game.
Here is the game with its basic treatment:
• Subjects are anonymously paired for a "one-shot" interaction with another
person.
• The role of “Proposer”, is randomly assigned to one of the subjects; the
other is then the “Responder”.
• The Proposer is given an endowment, the "pie" (e.g. $10), by the experimenters and the Responder knows the size of the pie.
• The Proposer then proposes how to divide the endowment between Proposer and Responder, transferring to the Responder any amount between
nothing and the entire endowment, e.g. the Proposer chooses to keep $ 8
and give $2 to the Responder.
95
96
MICROECONOMICS
- DRAFT
Player A
Player A
●
Offer (8,2)
split
Offer (5,5)
split
Player B ●
Accept
Offer (8,2)
split
●
Reject
Player A
●
Accept
Player B
Reject
●
●
●
●
(8, 2)
(0, 0)
(5, 5)
(0, 0)
●
Offer (5,5)
split
Player B
●
Reject
Accept
(8, 2)
(a) Full Game Tree
Offer (8, 2)
split
(0, 0)
Player B
Accept
Accept
(0, 0)
(8, 2)
(b) Self-regarding players
• If the Responder accepts the proposed division, the Responder gets the
proposed portion, and the Proposer keeps the rest.
• If the Responder rejects the offer both get nothing and the game ends.
Figure 2.10 presents a game tree for a variant of the Ultimatum Game where
the Proposer chooses between two offers: divide the pie equally and each
person gets $5 for an outcome (5, 5) or keep $8 and offer the Respondent
$2 for an outcome of (8, 2). The Responder then chooses whether to accept or reject the offer. The payoffs to each player are listed in the order of
play (Player A, Player B), so (8, 2) means Player A gets 8 and Player B gets
2.
If the Proposer cares only about her monetary payoffs in the game and believes that the Respondent is similarly self-regarding, then the Proposer
(Player A) will reason backwards as follows:
• Player A predicts that the Responder (Player B) will accept the offer of $2
(because A believes that B is also self-regarding and because $2 is greater
than his fallback of $0 which is what he gets if they reject the offer.
• A will propose the (8, 2) split to maximize A’s payoff.
• The Responder (B) will accept.
2.10 b 2.10 shows how the game unfolds if Player A and Player B are both
entirely self-regarding and maximize their monetary payoffs. Player B will
then always prefers a positive money amount over zero, and so they will
never reject a positive offer. Player A knows how Player B will respond, and
therefore, has a choice between a payoff of 8 and a payoff of 5; they prefer
8 and so offer a split of (8, 2). So if the two player’s are self-regarding then
backward induction leads to the Nash equilibrium of the game being (Offer (8,
2) Split, Accept) with payoffs (8, 2).
But A, the Proposer, who has followed through this reasoning now understands that she could have offered B the smallest possible positive amount,
say $1, and that the offer would have been accepted. Figure 2.10 c 2.10
Player B ●
Reject
(5, 5)
Offer (5, 5)
split
Player B
Reject
(0, 0)
Accept
(5, 5)
Reject
(0, 0)
(c) Reciprocal Player B
Figure 2.10: Game tree of the ultimatum
(bargaining) game. 2.10 a presents the full game
tree for both players regardless of type. Player A
is the Proposer and their actions are shown by the
blue branches. Player B is the Responder and their
actions are shown by the red branches.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
97
shows how the game might be played if Player B is a Reciprocator, that is,
Player B cares both about monetary payoffs and also about reciprocating how
Player A treats them. In this case, Player B views an offer of (8, 2) as unfair
or demonstrating bad intent, and they would rather get a payoff of zero dollars
than be treated poorly, so they would reject.
If, on the other hand, Player A offers (5, 5) then Player B views that as fair or
demonstrating good intent and they would prefer a payoff of 5 in that context
to a payoff of 0, so they would accept. Player A prefers a payoff of 5 to a
payoff of 0 (this is true regardless of whether player A is self-interested or
reciprocal) and so the Nash equilibrium of the game is (Offer a (5, 5) Split,
Accept) with payoffs (5, 5).
The Ultimatum Game has been played anonymously for real money in hundreds of experiments with university student subjects and other populations
– businessmen, fishermen, farmers, civil servants – in all parts of the world.7
The prediction based on the assumption that people are entirely self-regarding
and believe that others are too invariably fails as a description of how people
F AC T C H E C K Did the subjects not understand the game? It is not that complicated
a game, and later experiments in which
subjects played the game many times with
different partners showed this wasn’t true.
Their behavior remained consistent with
the one-shot experiments and their results
continued to be reproduced with many people making 50-50 splits (or nearly so) and
rejecting low offers.
behave. For example:
• Modal offers – the most common offers in the experiments – are typically
half of the pie, and average offers generally exceed 40 percent of the pie,
and
• Offers of 20 percent of the pie or less are often rejected with frequencies;
people in the position of Responder choose to reject and get zero rather
than accept and get a payoff of, say, $2 offered from the Proposers $10 pie.
These rejections of small but positive offers from the Proposer are interpreted
as evidence for reciprocity motives on the part of the Responder. Why? Because the Responder is willing to pay a price (giving up a positive payoff) to
punish the Proposer for making an unfair offer (an offer the Responder considers too low). Responders apparently consider a low offer to be a violation
of a norm of fairness, and a person with reciprocal preferences responds by
depriving the proposer of any payoffs at all.
Explaining the behavior of Proposers is more complicated. The outcomes of
the experiments are not sufficient to say whether the large number of even
splits (and other seemingly fair or near-fair offers) is explained by adherence
to fairness norms or altruism by the Proposer or to self-regarding preferences
informed by fear that the Responder will reject an unfair offer. The evidence
for reciprocity motives therefore, comes from the Responders’ behaviors, not
the Proposers’ behaviors.9
F AC T C H E C K Some have suggested that
the results were due to the relatively low
stakes in the game, such as the $10 mentioned earlier. But subsequent experiments
conducted among university students in
Indonesia for a ‘pie’ equal to three months
average expenditures replicated the results
as did experiments with U.S. students with
a ’pie’ ranging in size up to $100. Evidence
from France showed similar behavior by proposers with stakes ranging from 40 French
francs ($7.20) to 2000 French francs ($360)
(this was prior to the adoption of the Euro).
A further study in India observed stakes that
varied by a magnitude of over 1000: From 20
rupees ($0.41) to 20,000 rupees ($410) as
the stakes. 8
sparked curiosity among a group of behavioral scientists:
Was this simply an odd result, perhaps due to the unusual
circumstances of the experiment, or had Henrich tapped
real
differences, perhaps reflecting the distinct
98 behavioral
MICROECONOMICS - DRAFT
Figure 1.
provides some comparative information about the societies
discussed here. In selecting these, we included societies
both sufficiently similar to the Machiguenga to offer the
possibility of replicating the original Machiguenga results,
Locations of the 15 small-scale societies.
BEHAVIORAL AND BRAIN SCIENCES (2005) 28:6
2.10
A global view: Common patterns and cultural differences
Anthropologists and others were surprised that the results of experiments
with the Ultimatum Game have been so similar across the many countries in
which they have been conducted. One observed that in virtually all of the early
experiments the subjects were from WEIRD countries, meaning Western, Educated, Industrialized, Rich, and Democratic.10 A team of anthropologists and
economists (including one of your current authors) designed a series of experiments to explore whether the results reported so far are replicable in societies
with quite different cultures and social institutions and whether results differed
across the different societies.11 These societies included hunter-gathers,
herders, and farmers (some using modern methods, others not even having
cattle, horses, or plows). In their Ultimatum Game experiments the pie was
substantial, approximately a day’s average wages or other income.
Figure 2.11 shows the location of the 15 small-scale societies around the
globe. The team was wondering if they would find cultural differences, and
they found them.
Among the Au and Gnau people in Papua New Guinea offers of more than
half of the pie were common, and many of these high offers were rejected. In
fact Responders among the Au and Gnau peoples were as likely to reject a
offer of much more than half as an offer of much less than half.
799
Figure 2.11: Small-scale societies where the
Ultimatum Game experiments were conducted.
A map of the world showing the locations of the
small-scale societies where the Ultimatum Game
experiments were conducted. Source: Henrich
et al. (2005).
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
Though this seemed odd to the economists on the team, it did not surprise
the anthropologists who study New Guinea. They know that people in New
Guinea compete with each other to see who can give more or better gifts.
Gift-giving conveys status in their society and people use giving gifts as a
way to obtain status over others. Refusing a gift suggests that you are not
subordinate to the gift-giver, while accepting it means their status is higher
than yours.
By contrast, among the highly individualistic Machiguenga slash and burn
farmers in Amazonian Peru, almost three quarters of the offers were a quarter of the pie or less and there was just a single rejection, a pattern strikingly
different from other experiments. The Machiguenga came as close to acting like Homo economicus as any population yet studied. Even among the
Machiguenga, however, the mean offer was still 27 percent of the pie, more
than the zero we’d expect if they all were consistently self-interested.
The researchers who analyzed the experiments in the 15 small scale societies
made the following conclusions:
• Although behaviors vary greatly across societies, not a single society
approximated the behaviors that would be observed if everyone cared only
about their own payoffs and believed others were the same.
• Between-society differences in behavior seem to reflect differences in the
kinds of social interaction people experience in everyday life.
Here is some evidence that the experimental game behavior reflected the
lived experiences of the people.
• The Ache hunter gatherers in Paraguay share meat and honey equally
among all group members. Ache Proposers contributed half of the ‘pie’ or
more.
• Among the Lamalera whale hunters of Indonesia, who hunt in large crews
and divide their prey according to strict sharing rules, the average proposal
was 58 percent of the pie.
Given the evidence from small-scale societies like the Lamalera and the Ache,
we might ask whether we find other-regarding behavior in real-world situations
elsewhere in the industrialized world. A different team of researchers were
interested in exactly this question and designed an experiment that mirrors a
real life dilemma: what would you do if you found a wallet someone had lost:
would you return it?
The team distributed a total of 17,303 "lost" wallets some with money in them
some without, in 355 cities across 40 countries.12 In each country, the researchers targeted big cities to ensure that there was a good sample of subjects (and to ensure anonymity). Using transparent wallets with a business
99
MICROECONOMICS
Country
100
- DRAFT
Figure 2.12: Wallets with details of their owners
were more likely to be given back to their
owners when they contained money than when
they did not. The "Reporting Rate" is the fraction
of wallets that were "returned"
Switzerland
Norway
Netherlands
Denmark
Sweden
Poland
Czech Republic
New Zealand
Germany
France
Australia
Spain
Russia
Canada
Argentina
Israel
Portugal
USA
UK
Italy
Chile
Brazil
South Africa
Thailand
Mexico
India
Turkey
Ghana
Indonesia
Malaysia
Kenya
Kazakhstan
Morocco
Treatment
Money
No Money
20
40
60
80
Reporting Rate
card, grocery list, key and cash, the researchers could check how many people contacted the "owner" of the wallet given in the email address listed on the
business card to return the wallet.
Before reading on, ask what you think would happen in your community: how
many people would try to return the wallet? Would more people return the
wallet if it had money in it, than if it did not?
The results of people’s choices are shown in Figure 2.12. Though there were
differences across countries, with just one exception among the 33 countries
people were more likely to contact the "owner" if the wallet contained money
($13.45, the treatment) in it than if it did not ($0, the control). In a subset of
cases – in the US, UK, and Poland – the researchers added a treatment with
even more money in the wallet ($94.15). With a really substantial sum of
money in the wallet, people were as likely, if not more so, to contact the listed
email address on the business card in the wallet.
In interpreting the results keep in mind that the countries differ greatly in how
much an additional $13.45 would make to a person standard of living. Per
capita income in the richest countries in the sample (Norway for example) are
ten and even in some cases 20 times the per capita income in others (Kenya
for example), even when account is taken of the differing purchasing power of
each national currency at domestic prices.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
The evidence from both the Ultimatum Game and the wallet experiments
suggests two important take-aways:
• Culture matters: people from different parts of the world live by different
social norms and mutual expectations – what we can loosely call "culture."
People from different cultures differ in what they consider fair offers and
whether think it’s acceptable to make a self-regarding offer. They also differ
substantially in whether they will return a lost wallet.
• People are similar in many important respects: people across the world
have other regarding motives including altruism, fairness, and reciprocity.
In the "lost wallet" experiment in most countries a substantial fraction of
people attempted to return the wallet.
The Ultimatum game and the lost wallet experiment provide valuable information, but they are lacking in one respect. Most of the coordination problems
we face involve large numbers of people, like the people of the world making
decisions about their carbon footprint, or owners of businesses across the
entire economy deciding whether or not to invest, or the herders placing more
cows on the commons, or the people deciding whether to drive to work, or the
farmers of Palanpur deciding when to plant.
2.11
The Public Goods Game: Cooperation and punishment
A public good is one which more people can enjoy without reducing the
amount available to others, and from which others cannot be excluded from
access to the good. An example is global climate: it is experienced by everyone, and efforts to address the problem of climate change contribute to a
public good: that is a more sustainable environment. Another example is the
rules of calculus: if you learn how to differentiate that does not deprive others
of the knowledge of the same rules of differentiation.
This sounds like a good thing. But there is a problem. Why do people produce
or contribute to the provision of a public good? If nobody can be excluded
from enjoying the good, its hard to see how it would be possible to make
money by providing it. (Imagine trying to make a living by selling the rules of
calculus!)
We return to the problem of public goods in Chapter 5
It shares with the Prisoner’s Dilemma Game the feature that everyone could
do better if they agreed on a common course of action (i.e they all contribute)
but the dominant strategy for a self-regarding player is not to contribute. For
this reason a Public Goods Game is sometimes called an n-person Prisoners’
Dilemma because it has the same incentive structure.
101
102
MICROECONOMICS
- DRAFT
Rules of the Public Goods Game
The Public Goods Game experiment is designed to understand how people
will play in a game with this structure.
Here are the rules of the game:
• n players are each given an endowment of z.
• Each player simultaneously selects an amount ei , 0  ei  z to contribute
to the public good (think of ei as the player’s "effort" in contributing to the
public good).
• The amount of the public good produced depends on the level of contributions. For example it could be half of the sum of all of the contributions. In
this case the average productivity of contributions would be one-half.
• Each player, regardless of whether they contribute or not, obtain the entire
benefit of the total amount of the public good produced. ,
As a result of the rules, each player’s payoff can be read as follows:
Own payoff = Endowment
Contribution + Average productivity ⇥ Total Contributions
Figure 2.13 illustrates the benefits of the public good minus the costs of contributing to a public good in a 4-person public goods game. In the version of
the game we depict, they can each contribute $10 or $0: which we call "Contribute" or "Don’t." Now compare how a player does if they Contribute (red
line) or Don’t (blue line) if they are the only one who contributes, or there are
1, 2, or all 3 others contributing. You can see that in every case she will earn
higher payoffs by not contributing. Therefore, if all players are self-regarding,
the dominant strategy equilibrium is Pareto-inefficient and an alternative
outcome, full contribution by all, which is not a Nash equilibrium, is Paretoefficient.
A self-regarding player who cares about only their own costs of contributing and the benefits they get from the public good will choose to contribute
nothing (analogous to Defection in the Prisoners’ Dilemma). And this is the
case no matter how much or little the other players contribute. So contributing
nothing is the dominant strategy for each player. And, like in the Prisoners’
Dilemma, everyone contributing nothing is the dominant strategy equilibrium.
As a result economists expected that when this game is played for real money
that no player would contribute. They were in for a surprise.
M-Note 2.2: Why the dominant strategy in the Public Goods Game is
to contribute nothing
Payoff net of endowment, $
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
Figure 2.13: A 4-player public goods game with
choices to contribute or not. Each player can
play either Contribute or Don’t contribute, and as
there are 4 players, this means that the number of
others contributing can be any of the numbers 0,
1, 2, or 3 players playing either of the strategies.
Playing Don’t contribute yields a higher payoff for
the player regardless of how many players play
Contribute or Don’t. Therefore, Don’t contribute is a
strictly dominant strategy.
15
Contribute
Don't contribute
10
5
0
−5
0
1
2
3
Number of others playing Contribute
In Figure 2.13, there were four players and we limited their actions to either contributing
$10 or contributing $0, but in a standard public goods game players can contribute any
amount up to and including their entire endowment (such that e1 = z). In the public goods
game in which players can contribute any amount from their endowment, a player’s payoff
is given by Equation 2.2:
pi = z
103
ei + M Â e j for j = 1, . . . n
(2.2)
j
As earlier, we can break down this equation:
• z is the endowment of money the player receives from the experimenters.
• ei is the contribution a player makes at a cost to themselves.
• M is the multiplier, or the average productivity of contributions < 1.
• Â j e j is the total amount contributed by all players.
Whatever the other players do, you can see from Equation 2.2 if you differentiate p with
respect to hatei that for person i contributing one unit (say, penny) more changes the her
payoff by 1 + M which is the cost of contributing minus the public good that the contributor herself enjoys as the result of her contribution. So as long as M < 1 contributing
anything reduces the contributor’s payoffs. This is why not contributing is the dominant
strategy
Checkpoint 2.8: Two-action Public Goods Game
a. Draw a payoff table with two players, A and B, playing the Public Goods
Game. Limit their actions to e = 10 and e = 0 with M = 0.5. Check which is
the dominant strategy and explain why. What happens if M = 0.75?
b. Revise your payoff table and check what would happen if the strategies were
e = 1 and e = 0 with M = 0.5? Would anything change? What happens if
M = 0.75?
c. Think about the condition M < 1 < Mn. Why must this be true for the game
104
MICROECONOMICS
- DRAFT
Treatment
with punishment
without punishment
●
●
●
Average Contributions
12
●
●
●
●
●
10
●
●
●
8
●
●
●
6
●
●
●
●
●
4
●
●
2
●
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
Periods
to be an n-person Prisoners’ Dilemma game? (Hint: Think about what
would happen if it were not true. What would happen if M > 1? What would
happen if Mn < 1?)
2.12
Application: Evidence from Public Goods Games
The prediction of the self-regarding model that all players choose to contribute
nothing (e = 0) is consistently contradicted by the experimental evidence.
The evidence we have comes from people playing one-shot (single-period)
games and from people playing repeated games with as few as 5 rounds and
as many as 50 rounds.13
In one-shot games, contributions average about half of the endowment,
while in repeated games contributions begin around half and then decline
so that a majority of players contribute nothing in the final round of a ten-round
game.
Researchers have interpreted the decline in the first half as a reflection of
people getting disappointed about the expectations they had that other people would contribute more, along with the desire people have to punish low
contributors (or at least not to be taken advantage of) in a situation in which
one person can punish a low contributor only by reducing their own contributions.
In this interpretation it is the reciprocity motives of the higher contributing
subjects, disappointed or angry about their free-riding fellow subjects that
explains why cooperation unravels. So the decline in contributions becomes a
vicious circle: only by reducing how much they contribute can people punish
others, but in so doing other people might want to punish them for their low
contributions by contributing yet less again.
Figure 2.14: Public Goods Game with Punishment. Average contributions over periods 1 to
10 decrease without punishment. Over periods
11 to 20, subjects can be punished by their
peers and average contributions are higher on
average than in the first 10 rounds. Source:
Fehr and Gächter (2000a).
The vertical axis is the average contribution
each round. The horizontal axis is the period. At
period 11 the subjects are given the opportunity
punish each other. There are three treatments in
this Public Goods Game experiment. This figure
portrays the behavior in the "Strangers" treatment
where players are randomly re-matched each
round, but could have some players re-enter the
group. The two other treatments, which show
similar results, are "Partners" where players are
in the same group for all the rounds; and "Perfect
Strangers" where players are re-matched, but no
player will encounter any other more than once
during the experiment.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
The Public Goods with Punishment game
The idea of that the decline in contributions is due to the fact that in the standard game contributing less is the only way to punish low contributors is supported by an ingenious experiment. This has the same public goods structure
but with what turned out to be a major difference: after subjects contributed,
their contributions were then made public to all the group members, and members then had the opportunity to punish others in the group. Subjects could
punish them by imposing a cost, therefore, reducing the defectors total payoff.
In order to impose this cost, however, the Punisher also had to suffer a cost
themselves.
The change in the rules of the game – adding the punishment option – represents a change in the institutions governing contributions to the public good.
In the language of experiments the new rules are termed a new different treatment. So the standard game is one treatment and the game with punishment
is a second treatment.
In the experiment subjects engaged in extensive punishment of low contributors. At the start of the game people contributed over half of the endowment
and then, apparently in response to punishment of low contributors, they contributed more over the course of the game. The change in institutions modeled
by adding the punishment option altered the result dramatically.
To see if subjects’ willingness to punish could be based on the expectation
that they would benefit in subsequent rounds of the game, a slightly different
experiment was tried. The researchers adopted what they called a “perfect
strangers” treatment: after each round of the ten-round experiment the groups
were re-shuffled, so that no player ever encountered any other player more
than once. The "perfect strangers treatment" turned the experiment into a
series of one-shot games.
Since every player would only encounter every other player once, if lowcontributors responded to punishment by contributing more in subsequent
rounds, they would raise the payoffs of others but not the punisher (who would
never again be in the same group with the target of her punishment).
In this way, punishment itself became a public good. This is because a punisher incurs a cost, yet the benefits of punishing a low contributor and getting them to increase their contributions are non-rival and non-excludable.
Even in the perfect stranger treatment subjects avidly punished low contributors.
Further evidence comes from the fact that people punish low contributors
even in the last round of the game when punishment cannot be motivated
by the expectation that the punisher will benefit from their targets improved
behavior in the future. There is no future (the game ends after they punish).
105
106
MICROECONOMICS
- DRAFT
So reciprocal preferences – the pleasure of punishing a someone who is
violating a social norm – are most likely involved.14
Culture matters
When low contributors who are punished why do they subsequently contribute
more? You may think that the answer is obvious: they contribute more to avoid
the future reduced payoffs that being punished imposes on them.
But there must be something else going on. In two similar experiments – one
in the laboratory in the U.S. and one in the field among farmers in Zimbabwe
“punishment” was not in reduced payoffs, it was just a purely verbal expression of displeasure (e.g "selfish guy") by a fellow subject.15
But the targets of purely verbal punishment contributed more in subsequent
rounds. This occurs most likely because in many societies there is a norm that
people should contribute to the public good, and when a person is criticized
for violating it, they feel shame and try to make amends.16
Culture affects experimental play in other ways. The anthropologist Jean
Ensminger conducted public goods experiments with the Orma, a herding
people in Kenya. Members of the Orma regularly voluntarily contribute their
labor to producing some public good – for example, the repair of a road –
a system they call in Swahili, "Harambee." Families that have more cattle –
more wealth – are expected to contribute more to the project.
When Ensminger explained the Public Goods Game to the Orma participants,
they promptly called it the “Harambee game.” Those with more cattle contributed more in the experiment, just as would have been the case in a real
"Harambee."
When the Orma subjects played the Ultimatum Game, however, they did
not compare it to "Harambee" and wealth did not predict any aspect of their
experimental play.17 This difference in how the wealthy played the Public
Goods game and the Ultimatum Game probably would not have surfaced
among the farmers you have already met from, Palanpur, Illinois or West
Bengal.
2.13
Social preferences are not "Irrational"
People sometimes think of other-regarding and ethical preferences as something special – different from the taste for ice cream, for example – and requiring a model different from the preferences, beliefs, and constraints approach.
But the desire to contribute, to punish those who do not contribute, and otherwise to act on the basis of social preferences, like the desire to consume
conventional goods and services, can be represented by preferences that
conform to standard definitions of rationality.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
107
What we know from experiments is that whether its ice cream or contributions
to the public good, people respond to trade-offs, taking account of the costs
of and how much they value the activity in question: the higher the cost of
helping others, the less its frequency. In other words, other-regarding preferences are consistent with rationality, namely consistency (transitivity) and
completeness.
Researchers tested the rationality of seemingly altruistic choices by asking
176 subjects to play a version of what is called the Dictator Game.18 In what
is called the “Dictator Game", one player (the Dictator), Alice, is given a sum
of money by the experimenter, and asked to transfer whatever proportion of
the money that she wishes to an other (anonymous) subject, Bob. Alice is
told that that for every dollar that Bob receives from her, she will have to pay p
dollars. So p is the price of altruism:how much she has to pay for every dollar
that Bob gets. After Alice makes her decision, the money is transferred, and
the game is over.
In this experiment, 75 percent of the Dictators gave away some money,
demonstrating altruistic preferences. The average amount given away was
a quarter of the endowment when the price p = 1 (a dollar for-dollar transfer).19 Moreover, the higher the price of generosity, the less money was transferred. For instance, when each dollar transferred to Bob cost Alice two dollars
( p = 2), only 14.1 percent of the endowment was given away on average, and
when each dollar transferred cost four dollars, only 3.4 percent of the dictator’s
endowment was transferred. The higher the price of altruism, the less did
Alice "purchase."
It may be, as the old saying goes, that "virtue is its own reward." But that
does not mean that people will act virtuously no matter what the price. This
finding is perfectly consistent with the fact that people respond to the price
F AC T C H E C K In a Public Goods Game with
Punishment experiment researchers found
that the level of punishment that subjects
inflicted on others was less when each dollar
subtracted from the payoffs of the target cost
more in foregone payoffs to the punisher.20
of virtuous behavior just as the preferences, beliefs, and constraints model
predicts.
Checkpoint 2.9: Dictator Game?
Is the "Dictator Game" a game? Think about how we’ve defined games (check
back in Chapter 1 to ensure you remember).
2.14
Application: The lab and the street
Do people behave in the real world the way they do in experiments? The
experimental evidence for reciprocity or related forms of other-regarding
behavior would not be interesting if it did matched by similar behavior outside
the lab. We therefore need to check whether laboratory evidence is externally
valid, that is, consistent with behavior observed outside of the laboratory in
similar circumstances to those found in the lab. External validity is particularly
E XTERNAL VALIDITYResults of experiments
or other scientific research that can be
generalized to circumstances outside
(external to) the laboratory or other setting in
which the research was produced, are said
to be externally valid.
108
MICROECONOMICS
- DRAFT
important for policy questions because policy-makers and governments need
to know whether a policy will work outside of the controlled conditions of the
laboratory.
Generalizing directly from experiments to behavior in other contexts is often
unwarranted. For example, in the Dictator Game typically more than 60 percent of the Dictators allocate a positive sum to the recipient, and the average
given is about a fifth of the endowment.21 But we would be sadly mistaken if
we predicted on the basis of this experimental result that 60 percent of people
would spontaneously give money to an anonymous person passing them on
the street, or that the same subjects would offer a fifth of the money in their
wallet to a homeless person asking for help.
Many researchers have tried to see whether behavior in lab experiments
predicts behavior outside the lab.
Along the coast of northeastern Brazil, for example, shrimpers catch shrimp in
large plastic bucket-like contraptions. The shrimpers cut holes in the bottoms
of the traps to allow the baby shrimp to escape, thereby preserving the stock
of shrimp for future catches.
The shrimpers face a real-world coordination problem: the expected income
of each would be greatest if he were to cut smaller holes in his traps (increasing his own catch) while others cut larger holes in theirs (preserving future
stocks). In Prisoners’ Dilemma terms, small trap holes are a form of defection
that maximizes the individual’s material payoff irrespective of what others do (it
is the dominant strategy). But a shrimper might resist the temptation to defect
if he were both public spirited toward the other fishers and sufficiently patient
to value the future opportunities that they all would lose were he to use traps
with smaller holes.
Ernst Fehr and Andreas Leibbrandt implemented both a Public Goods game
and an experimental measure of impatience with the shrimpers. They found
that the shrimpers with both greater patience and greater cooperativeness
in the experimental game punched significantly larger holes in their traps,
thereby protecting future stocks for the entire community.23
Additional evidence of external validity comes from a set of experiments and
field studies with 49 groups of herders of the Bale Oromo people in Ethiopia,
who were engaged in forest-commons management. Devesh Rustagi and
his coauthors implemented public-goods experiments with a total of 679
herders, and also studied the success of the herders’ cooperative forest
projects.24
The most common behavioral type in their experiments, constituting just over
a third of the subjects, were reciprocators who responded to higher contributions by others by contributing more to the public good themselves. The
F AC T C H E C K In an experimental game
about trust and reciprocity played by groups
of students and groups of chief executive
officers of Costa Rican businesses, the
businessmen were both more trusting of
others and also reciprocated the generosity
of their game partners to a far greater degree
than did the students.22 Based on existing
experimental evidence, students are not
particularly other-regarding.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
109
authors found that groups with a larger number of reciprocators were more
successful – they planted more new trees – than those with fewer reciprocators. This was in part because members of groups with more reciprocators
spent significantly more time monitoring others’ use of the forest. As with the
Brazilian shrimpers, differences in the fraction of reciprocators in a group were
associated with substantial increases in trees planted or time spent monitoring
others.
2.15
Application: A fine is a price
Figure 2.15: A shrimping bucket with holes in it
How might a policy-maker or CEO of a business make use of the fact that
people care about what happens to others and they value behaving ethically?
Think about a set of rules for compensating employees. The rules typically
specify pay and provision for time off, sick days and the like. But problems
arise with using purely material incentives to influence how people behave.
Having noticed a suspicious bunching of sick call-ins on Mondays and Fridays,
the Boston Fire Commissioner on December 1, 2001 ended the Department’s
policy of unlimited paid sick days. Instead, the commissioner imposed a 15day sick day limit. The pay of firefighters exceeding that limit would be cut.
The firefighters responded to the new incentives: those calling in sick on
Christmas and New Year’s Day increased ten times over the previous year’s
sick days.
The Fire Commissioner retaliated by cancelling their holiday bonus checks.
The firefighters were unimpressed: the next year they claimed 13,431 sick
days; up from 6,432 the previous year.25
Many of the firefighters, apparently insulted by the new system, abused it,
or abandoned their previous ethic of serving the public even when injured
or not feeling well. In the language of the Ultimatum Game, they responded
reciprocally to an offer they disliked by rejecting it. They were trying to punish
the Commissioner at a cost to themselves.
The Commissioner’s difficulties are far from exceptional.
Consider the following experiment in Haifa, Israel.26 Parents everywhere are
sometimes late in picking up their children at day care centers.
• Treatment: At six randomly chosen day care centers, a fine was imposed
for parents picking up their children late.
• Control: In a control group of day care centers no fine was imposed.
110
MICROECONOMICS
- DRAFT
Fines
Group with fine
Control group
20
●
●
●
●
●
●
●
Late Arrivals
●
●
●
18
●
●
Figure 2.16: The effect of a fine for lateness in
Haifa’s day care centers. Source: Gneezy and
Rustichini (2000a). The fine was imposed in week
5 and retracted in week 17.
●
16
●
14
●
12
●
●
●
●
●
10
●
●
●
●
8
●
●
●
●
●
●
●
●
●
1
2
3
4
●
●
●
●
6
●
●
5
●
●
●
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20
Week Number
Researchers expected parents to arrive on time because of the fine. But parents responded to the fine by arriving late more often: the fraction of parents
picking up their kids late more than doubled. when the fine was taken away
after 16 weeks, the parents continued to arrive late, showing no tendency to
return to the status quo prior to the experiment. Over the entire 20 weeks of
the experiment, there were no changes in the degree of lateness at the day
care centers in the control group.
The researchers reason that the fine was a contextual cue, unintentionally
providing information about appropriate behavior. The effect was to convert
lateness from the violation of a social norm or obligation that the parents were
to respect, to a choice with a price that many were willing to pay. They titled
their study “A Fine is a Price” and concluded that imposing a fine labeled the
interaction as a market-like situation, one in which parents were more than
willing to buy lateness for money. Revoking the fine did not restore the initial
context.
When monetary incentives undermine social preferences as they did among
the Boston firefighters and Haifa parents, this is called crowding out. These
two cases of crowding out are cautions that the use of monetary incentives
may be inappropriate where the targets of the incentives are motivated by
other regarding preferences. But they are not reasons to think that incentives
are ineffective, as we will see in many examples to follow. We have no doubt
that had the fine for lateness in Haifa been 500 New Israeli Shekels rather
than 10 the parents would have found a way to pick up their kids on time.
C ROWDING OUT is said to occur when monetary or other material incentives undermine
other regarding or ethical preferences.
2.16
Complexity: diverse, versatile, and changeable people
The experimental and observational evidence suggests an adequate understanding of preferences should recognize four aspects in human social
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
111
behavior.
• Diversity : people differ in their preferences .
• Versatility : even a single person has a diversity of preferences, and which
of these is salient for making a decision depends on the situation, for
example, shopping as opposed to spending time with friends.
• Changeability : people learn new preferences under the influence of their
experiences.
These three aspects of our preferences contribute to a fourth attribute of how
human beings interact:
• Complexity or "the whole is not the sum of its parts": the outcome of an
interaction of many people cannot be deduced in any simple way from the
characteristics of the individual people involved.
Diversity
What motivates people differs, both locally and across different cultures and
across time. Using data from a wide range of experiments, researchers estimate that between 40 and 65 percent of people exhibit other-regarding
preferences of some kind. The same studies suggest that between 20 and 35
D IVERSITY AND H ETEROGENEITY We use
the term heterogeneous to describe a group
of actors with different preferences or some
other attributes, for example, wealth gender,
nationality, first-mover status in game, and so
on.
percent of the subjects exhibit conventional self-regarding preferences.27 The
authors of another study (in the U.S.) termed 29 percent of their experimental
subjects as "ruthless competitors" (presumably resembling Economic Man)
and 22 per cent as "saints."28
Versatility
A common observation about human behavior made by psychologists is
that the same person can act differently depending on the situation. As a
result, we say that people are versatile: they change in response to what their
situation seems to require of them, for example, being self regarding while
shopping and other regarding with one’s neighbors.
In the Ultimatum Game, Proposers often offer amounts which maximize their
expected payoffs. But Responders rarely do. Researchers have also found
this in experiments where subjects play both roles: Proposer at one stage
in the experiment and Responder in another stage. The same person when
in the role of Responder typically rejects positive offers if they appear to be
unfair, even if they had made a similar low offer when in the role of Proposer.
They therefore act as if they had reciprocal or inequality averse preferences
as Responders, despite exhibiting self-regarding preferences when they are
Proposers. The fact that in the role of proposers people are more like "ruthless
competitors" while in the role of responder are more like "saints" is evidence of
our versatility.
V ERSATILITY How people act depends
very much on the situation, resembling
Homo economicus in some contexts (say,
in business), but other-regarding social
preferences in other contexts (say, around
their family). Psychologists explain how
a situation can frame a decision so as
to suggest appropriate attitudes (or as
economists would say, preferences) towards
the possible actions an individual might
make. We refer to this aspect of our behavior
as versatility.
112
MICROECONOMICS
- DRAFT
Changeability
Some preferences are part of our genetic makeup, having a taste for sweet
and fatty foods, for example. But most preferences are learned rather than
given by our genetic inheritance. Durable changes in an individual’s evalua-
F AC T C H E C K In experimental games
about dishonesty, people who grew up in
Communist Party ruled East Germany are
more likely to cheat than those who grew up
in West Germany.29
tions of outcomes often take place as a result of experience. When this occurs
we say that preferences are endogenous, meaning that they change as a result of influences such as where a person lives, how they make their living or
the rules of the game that govern how they interact with others.
Over a lifetime or even generations, migrants to a new country, or those moving from a rural to an urban area often adopt new preferences (for example
concerning food tastes). The fact that preferences are learned may account
for the fact that, as we saw from the experiments in small scale societies, people who hunt large animals tend be generous with the meat they acquire; and
they seem to generalize these habits to other realms of life. Preferences are
exogenous of they do not change or change only in response to influences
that considered to be external.
A consequence: Complexity
In everyday language the word "complexity" refers to the state of being intricate or complicated. The term is used in quite a different way in the study of
interactions of a large number of independent entities – whether particles or
people. A key idea is that the results of these interactions for the system as
a whole cannot be predicted in any simple way from even the most detailed
knowledge of the interacting entities.
The best example of complexity in the social sciences is Adam Smith’s invisible hand. What Smith suggested two and a half centuries ago, and modern
economics has shown (as seen in Chapter 15) is that under some conditions uncoordinated interactions among entirely self-regarding total strangers
through competition in markets among private property owners can (unwittingly) create an outcome that is better for all than many of the alternatives.
The idea of complexity is often expressed the adage: the whole is different
from the sum of the parts. The key here is not that the whole may be greater
or less than the sum; it is that summing the parts is not the right way to calculate the whole. Averaging the components of some interacting system will not
give what their interactions will actually add up to. The results of the interaction – called their emergent property – may be surprising given the nature of
the interacting entities.
Here are some examples of surprises (with which you are already familiar)
in the properties that emerge from people with heterogeneous and versatile
preferences interacting.
E NDOGENOUS PREFERENCESPreferences
are endogenous if they change as a result
of influences such as where a person lives,
how they make their living or the rules of
the game that govern how they interact with
others.
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
• Small differences in the distribution of types – the presence in a population
of a few people willing to punish those who do not contribute in a Public
Goods Game, for example – can have large effects on how everyone behaves, getting the self-regarding people to act as if they were cooperators.
You have seen this in Figure 2.14.
• Seemingly small differences in institutions can make large and surprising
differences in outcomes. Why did adding the punishment option so radically change the outcomes in the Public Goods Game? We know that
cooperation – contributing to the public good – unravels in in the absence
of the punishment option. But the incentives to punish would seem identical
to the incentives to contribute to the public good in the first place: everyone
would like someone else to bear the cost of punishing the free riders. So
not contributing and not punishing should be the dominant strategy in this
game. But we now know that that is not what we observe.
• While imposing a fine or other cost on socially undesirable behaviors may
create socially desirable outcomes in certain circumstances such as getting
people to stop using plastic grocery bags , a fine on parents arriving late
to pick up their kids backfired. We saw that the nominal fine decreased
parents’ willingness to pick up their children on time when the viewed the
fine as a price to pay for additional day care: the fine changed what they
viewed as socially acceptable behavior.
• Letting a self-regarding player be the first mover in a Prisoners Dilemma
game when she knows that the other player has strong reciprocity motives
can avert the coordination failure resulting in mutual cooperation. Letting
the Reciprocator be the first mover would have the opposite result: both
players would defect, resulting in the Pareto inefficient outcome. You can
confirm this by doing Checkpoint 28
Checkpoint 2.10: Sequential Prisoners’ Dilemma: Self-interest versus
reciprocity
For a sequential Prisoners’ Dilemma game where the first player is selfinterested and the second player is reciprocal draw a game tree in which the
Nash equilibrium may be (Cooperate, Cooperate) and explain why could occurs.
2.17
Conclusion
We have explored various social interactions represented as games and also
as studied empirically using experiments, such as the Ultimatum game and
the Public Goods Game. While self-regarding preferences are represented an
essential and powerful motivator for human behavior, we have also found that
113
114
MICROECONOMICS
- DRAFT
people behave cooperatively – viewing it as the right thing to do – and they
enjoy behaving cooperatively.
People dislike unfair treatment and enjoy punishing those who violate norms
of fairness or cooperation. The evidence that social preferences are common
does not, however, suggest that people are irrational. Indeed, as we have
seen, the experimental evidence suggests strongly that when individuals
give to others their behavior conforms to the requirements of rational choice.
People respond to the price of giving, giving more when it costs them less to
benefit the people who receive their money.
The importance of other-regarding preferences thus, does not challenge the
assumption of rationality or the preferences, beliefs and constraints approach.
However, it does suggest that for many applications we should take account of
people’s concerns for others and for doing the right thing.
Making connections
Preferences, beliefs, and constraints: This framework for analyzing decisions
will be used throughout the rest of the book.
Risk and uncertainty: Many, maybe most, of the important decisions that
people make are risky because the resulting outcome depends on contingencies the probability of which occurring the actor does not know.
The rules of the game and coordination problems: Sequential rather than
simultaneous play may result in a better outcome in an Assurance Game
(or even a Prisoners Dilemma). The reason is that the first mover can
help to coordinate play in the game. Another example: allowing players to
punish low contributors in a public goods game dramatically changes the
outcome.
External effects and Pareto-inefficient Nash equilibria: The public goods
game illustrates an extreme form of positive external effects (each person’s
contribution benefits everyone equally).
Evidence: Economists have recruited novel experimental evidence – from the
laboratory and the field – to examine our theories about how people behave. Economists have used the evidence to modify and improve existing
models and to develop entirely new models of how people behave.
Heterogeneity: People differ in their preferences (self-regarding, other regarding) and in the advantages associated with their positions (first mover,
second mover)
PEOPLE: SELF-INTEREST AND SOCIAL PREFERENCES
Important ideas
preferences
beliefs
constraints
rationality
self-interest
social preferences
fairness
altruism
reciprocity
spite
endogenous (preference)
exogenous (preference)
institutions
ad hoc
external validity
laboratory experiment
field experiment
endowment
ultimatum game
public goods game
punishment
group membership
complexity
inequality aversion
diversity
changeability
learning
Mathematical Notation
Notation
Definition
x
P
p ()
E (), p̂
z
e
M
n
p
a contingency
probability of a contingency to occur
a player’s payoff
a player’s expected payoff
individual endowment in Public Goods Game
individual contribution in Public Goods Game
return factor to contributions in Public Goods Game
number of participants in Public Goods Game
price of altruism in Dictator Game
Note on super- and subscripts: i: an individual; Early: the strategy planting
early; Late: the strategy planting late.
Discussion questions
See supplementary materials.
Problems
See supplementary materials.
115
3
Doing the best you can: Constrained optimization
DOING ECONOMICS
“What a useful thing a pocket-map is!” I remarked.
“That’s another thing we’ve learned from your Nation,” replied Mein Herr, “mapmaking. But we’ve carried it much further than you."
“What do you consider the largest map that would be really useful?”
“About six inches to the mile.”
“Only six inches! “ he exclaimed.
“We very soon got to six yards to the mile. Then we tried a hundred yards to the
mile. And then came the grandest idea of all! We actually made a map of the
country, on the scale of a mile to the mile!”
“Have you used it much?” I enquired.
“It has never been spread out, yet,”
“The farmers objected: they said it would cover the whole country and shut out
the sunlight!"
Lewis Carroll, Sylvie and Bruno Concluded, 1893
Lewis Carroll, the author of this dialogue (not to mention Alice in Wonderland)
was also a mathematician and a philosopher. The point Carroll made about
maps also goes for economic models. Maps are useful because they convey
the necessary information, not because they are an exact representation of
the territory, as the people from Mein Herr’s country discovered. Carroll’s
point? The map is not the territory.
A good model is not reality, but it’s a helpful guide.
This chapter will enable you to:
• See how the preferences, beliefs and
constraints framework from Chapter 2
forms the basis for mathematical models
of economic behavior.
• Recognize how preferences – whether
entirely self regarding or altruistic – can
be represented both in mathematical
form (a utility function) and graphical
form (an indifference curve map).
• Understand that constrained optimization
is a method that economists use to
explain the actions that people take; it
is not a description of the thoughts or
feelings making up individuals’ decision
making processes (e.g.studied by a
psychologists).
• Explain how people are constrained – for
example by limited time – and how these
constraints give rise to opportunity costs
and, along with our preferences, to trade
offs.
• Use the preferences, beliefs and constraints framework to analyze difficult
choices concerning in policy-making, including how much of society’s resources
should be devoted to the abatement of
environmental damages.
• Use the concepts of ordinal and cardinal utility explain how they differ and
how cardinal utility provides a way to
represent the societal cost of economic
inequality.
• Understand the shortcomings and limits
as well as the insights of these models.
What qualifies a map or a model as useful depends on what we need it for:
six inches to the mile might be adequate for a map of hiking trails, but such a
hiking map would not be much use to an airplane pilot. The same is true of
economic models.
Figure 3.1: The London underground transit
system: the map represents, but is not the same
as, the territory to which it corresponds. The map is
a helpful model.
118
MICROECONOMICS
- DRAFT
Think of a model as a lens. An economic model is a way of focusing on what
is important given the question that one wants to address without complicating
the picture with things that do not matter for the question at hand.
A key component of many economic models – those using the preferences,
beliefs and constraints approach – is that we can understand the actions
people take by assuming that they are doing the best they can under the
circumstances that they are in. When implemented using mathematical
reasoning, this process is called "constrained optimization, a process by which
a person determines a course of action to accomplish a goal (reflecting the
person’s preferences), given the information that the person has (beliefs) and
the actions they may feasibly takes (a constraint).
C ONSTRAINED OPTIMIZATION is the mathematical representation of a process by which
a person determines a course of action in
order to accomplish a goal (reflecting the
person’s preferences), given the information
that the person has (beliefs) and the actions
they may feasibly takes (a constraint.)
We illustrate a model and the process of constrained optimization by something that matters to all of us: Time, and how we use it.
3.1 Time: A scarce resource
Benjamin Franklin (1706-1790) – the American politician and inventor – once
said, "Time is money." Franklin was referring to the presence of trade-offs in
how people choose to spend their limited time. His three-word sentence is
therefore a constrained optimization model: people choose their daily actions
to achieve their goals under the constraint of limited time.
Spending an hour or minute on an activity provides us value of some kind:
we enjoy the activity itself (e.g. eating) or the results of the activity (e.g. being
paid a wage). But, since time is limited, choosing one activity also means we
give up that time to do something else. We incur a cost of doing an activity
because we forfeit the value of the next best thing we could have spent our
time on instead: this is the opportunity cost of our time.
Unless we have time to spare, and are wondering how we will fill up our day
there is an opportunity cost to our use of time. As a result, we can model
how we use our time as the result of our evaluating the benefits and costs
(including opportunity costs) of pursuing one set of activities rather another. To
do this we use constrained optimization.
Before developing the concepts on which constrained optimization is based,
let’s look at the kinds of facts that a model of time use should be able to explain.
Figure 3.2 shows how men and women from the USA used their time each
day during the year 2013. The largest time use is for the categories sleep,
work (meaning for pay), leisure, and house work. Men and women differ
typically in the hours they devote to paid work and house work and care work,
often reflecting differing social norms about the kinds of activities that it is
"appropriate" or "natural" for men and women to do.
O PPORTUNITY C OST The opportunity cost
of x in terms of y is the marginal rate of
transformation: how much y a person must
give up to get a unit more of x .
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
119
Figure 3.2: Daily time use of American men and
women. These data – for hours in each activity
measured on the horizontal axis for all adults
– differ from data restricted to those with small
children, or retired people, or students.
8
Hours
6
Gender
Men
4
Women
2
pi
ng
ee
k
Sl
W
or
d
or
ts
Sp
&
e
Le
is
ur
Pa
i
k
ew
or
k
ou
s
e
w
or
H
D
&
ar
rin
ki
lC
Ea
tin
g
C
e
ar
ng
so
na
op
pi
Pe
r
Sh
ng
0
Source: Hofferth, Flood, and Sobek (2013).
But these social norms also change, sometimes in ways that show that the
differences in the distribution of work time between men and women are far
from determined by "nature" but instead reflect changed economic conditions.
During the second half of the 20th century in the rich countries the fraction of
women doing paid work outside the home dramatically increased. While we
do not have detailed information like that shown in Figure 3.2 for the mid-20th
century on how men and women spent their time there almost certainly has
been a decline in the amount of time doing housework.
Part of the change in the distribution of women’s time between house work
and work for pay is due to the availability at affordable prices of new technologies – household appliances – that reduced the amount of time required
to clean house, wash clothes, and the carry out the other housework tasks.
These include washers, refrigerators, and vacuum cleaners which in the U.S.
became common from the late 1940s onward, and dryers, dishwashers and
microwaves somewhat later.
Evidence that these new technologies contributed to the change in the distribution of women’s work time comes from a comparison across countries
of increases in the fraction of women working outside the home – called the
labor force participation rate – and decreases in price of these labor saving
household appliances (compared to other prices).1 The results are in Figure
3.3, which shows that in countries such as the U.S. where the prices of these
appliances fell the most, women’s labor force participation rate rose the most.
By contrast, in Germany where prices of household appliances fell the least,
the increase in labor force participation was half as great as in the U.S.
Other factors contributed, of course, most importantly the reduction in the
F AC T C H E C K For a long historical view of
why the washing machine was a "miracle"
have a look at this video by Hans Rosling:
<LINK HERE>
- DRAFT
MICROECONOMICS
Change in Female Labor Force Participation
120
15
Figure 3.3: The relative price of home appliances and the female and male labor force
participation rates. The vertical axis represents
an index that records the change in the fraction of
adult women working outside the home, termed
the female labor force participation rate (FLFP),
as well as the change in the home appliance price
index (HAPI) on the horizontal index. The figures
shows that as the price of labor saving household
appliances decreases, the female labor force participation rate increases. Notice that a bigger price
decrease would be shown by a larger negative
change (further to the left on the x-axis) so the US,
Denmark and the Netherlands had big decreases
in the prices of home appliances and a big increase
in the female labor force participation rate. Household appliances – like TVs – that did not reduce the
amount of time necessary to perform housework
tasks are excluded. Source: de V. Cavalcanti and
Tavares (2008).
Netherlands
14
13
12
11
US
Denmark
10
UK
9
Luxembourg
8
Italy
Belgium
Ireland
France
7
Germany
6
5
−0.30
−0.25
−0.20
−0.15
−0.10
Changes in the Home Appliance Price Index
number of children born per woman. But the fall in the prices of appliance,
the study concluded, was of approximately equal importance. It appears that
economic changes – the new household appliances and their falling prices
– changed how women spent their time – more working outside the home .
This in turn may have been a both a result and a cause for the changing social
norms about "women’s work" and the decreased adherence to the ideal of
a family with a husband income earner and a wife raising (many) children
and taking care of the home. This is an example of how preferences – for
example, these norms – change as economic conditions – the prices of home
appliances – change.
We begin with these examples because methods of constrained optimization
– the preferences, beliefs, and constraints framework – provide a way of
posing and in some cases answering questions like: Why do men and women
spend the time they do on the various activities shown? Or why did work
hours fall so dramatically in some countries over the 20th century?
We begin with with preferences, before turning to constraints later in the chapter. Because we are not considering strategic interactions or other situations
in which the relevant facts are not known, we do not treat beliefs until the next
chapter.
Checkpoint 3.1: Labor-saving household appliances and women’s labor force participation
Imagine a conversation around the year 1970 between a husband and wife who
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
121
just learned that a very effective clothes washing machine is available at a low
price. How might the conversation have led to the woman taking up paid work
outside the home?
3.2 Utility functions and preferences
In Chapter 1, we represented preferences – our evaluations of the outcomes
our actions may bring about – as payoffs that is, numbers indicating how much
the decision maker values each of the possible outcomes. We discussed, as
an illustration, the choice of whether to take an umbrella or not, with a decision
[Don’t take the umbrella, It rains] resulting in a payoff of 3. The payoff to [Take
the umbrella, It rains] was 15, meaning that if it rains the person valued having
the umbrella by 5 times as much as not having it.
In that example we simplified things by limiting the actions and the outcomes
to just a few, for example, it either rained or it did not. The simplification
allowed us to focus on 2 by 2 payoff matrices with just four possible outcomes.
But most of the economic interactions that we study are not that simple:
we can contribute any amount to the public good (not just $10 or nothing),
the farmers in Palanpur have the choice to plant a little bit earlier, or much
earlier, and so on. Or, to return to the question of time: how we divide up
our day among the activities in Figure 3.2 could be measured in variations of
minutes devoted to each of the nine activities, giving us trillions of "outcomes"
to choose amongst. We need a way of representing preferences when there
are a great many outcomes, without expanding our payoff matrices to the
unusable size of the 1:1 maps in the Lewis Carroll fable at the beginning of the
chapter.
Why we use utility functions to represent preferences
To do this we use a utility function, a mathematical expression that translates the full range of possible outcomes into a person’s valuation of the outcome – her payoffs.
The word "utility" (in ordinary language, "usefulness") is used to mean the
same thing as "payoff." It is a number assigned to a particular outcome bundle
that has the property that when choosing between alternative bundles, a
person will select the one with the highest (utility) number.
Both "utility" and "payoff" sound like some monetary or other amount of something you take home as the outcome of a game. But in economics utilities, like
payoffs are not something you get or even experience. You don’t take them
home, they are nothing more than numbers that indicate the course of action
you will take.
R E M I N D E R : P R E F E R E N C E S represent the
favorable (positive) or unfavorable (negative)
feelings that could lead a person to choose
one outcome over another. Included are
tastes (food likes and dislikes, for example),
habits (or even addictions), emotions (such
as anger and disgust) often associated
with visceral reactions (such as nausea or
an elevated heart rate), social norms (for
example, those that induce people to prefer
to be honest or fair), and psychological
tendencies (for aggression, extroversion, and
the like). Do not think about a preference or
the number a utility function assigns to some
bundle as "how much a person likes" the
bundle.
U TILITY FUNCTION A utility function is
an assignment of a number u(x, y), to
every outcome bundle (x, y) representing
a person’s valuation of that bundle. This
means that if given the choice between two
bundles (x, y) and (x0 , y0 ), the individual will
choose the first if u(x, y) > u(x0 , y0 ).
122
MICROECONOMICS
- DRAFT
For simplicity, we call this number "how much the person values the outcome"
but the utility function tells us nothing about why the bundle has a higher number. It could be any of the reasons for the collection of pro or con evaluations
that make up our preferences for some bundle, ranging from food tastes to
addictions to ethical norms.
What the function allows us to do is to take account of more complex outcomes than "Don’t take the umbrella" and "It rains." The decision maker, as
before, will choose the actions the she believes will result in the highest utility outcome. Suppose that our decision-maker, Annette, an Uber driver, is
deciding how much time to work, x, and what fraction of the resulting income
to spend on food, y. The utility function then assigns a number – the level of
utility – to each possible combination of x and y, say, work for 4 hours and 15
0
00
M - C H E C K We read x as ’x prime" and x
as "x double prime". We usually denote a
bundle other than (x, y) as (x0 , y0 ) to indicate
a different composition of the underlying x
and y.
minutes and spend 35 percent of the resulting pay on food. Any other combination, say, work four hours and spend 40 percent of the resulting income on
food, will be assigned another number, representing Annette’s valuation of that
particular outcome.
This assignment of numbers is a utility function, u(x, y): for every outcome
(x, y) the value of the utility function is the number representing a person’s
valuation of the outcome. Then if we know what combinations of x and y are
available to Annette based on the relevant constraints, then we can predict
the choice Annette will make, namely the combination with the highest utility.
What do the utility numbers measure?
We measure how much a person values various outcomes in two ways, either:
• by indicating how valuable each is on some absolute scale, or
• by simply ranking them in order.
If Annette compares two bundles (or outcomes), namely (x, y) and (x0 , y0 ) with
u(x, y) = 3 and u(x0 , y0 ) = 9 there are two statements we could make about
Annette, one much more informative than the other:
• Annette values (x0 , y0 ) three times as much as (x, y) and
• Annette values (x0 , y0 ) more than (x, y)
In the first case above, utility is a number indicating by how much Annette
prefers (x0 , y0 ) to (x, y). Utility is therefore termed a cardinal measure (cardinality in mathematics refers to the size of something). In Chapter 2 we
represented people’s preferences by the payoffs associated with particular
outcome bundle of games like (x0 , y0 ) or (x, y). When we defined the expected
payoffs to some course of action we added up the payoffs of each possible
C ONSISTENCY Consistency (or transitivity)
requires that when considering three bundles
(x, y), (x0 , y0 ), and (x00 , y00 ), if (x, y) is
preferred to (x0 , y0 ) and (x0 , y0 ) is preferred
to (x00 , y00 ), then (x00 , y00 ) cannot be preferred
to (x, y). Consistent preferences can
never lead someone to make contradictory
choices.
C OMPLETENESSCompleteness requires that
all possible outcomes can be ranked. For
any two bundles (x, y) and (x0 , y0 ) either
the person prefers (x, y) to (x0 , y0 )) or the
person prefers (x0 , y0 ) to (x, y)) or the person
is indifferent between (x0 , y0 ) and (x, y)).
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
123
outcome (weighting them by the probability of each outcome occurring). Doing this required that utility is a measure of size. The numbers representing
payoffs in Chapter 2 are cardinal utilities In the second case the utility function gives us an ordering of better-worse for the pair of outcomes. When the
utility function is measured in this way, we say that Annette has ordinal preferences or that utility is ordinally measured. Ordinal utility says nothing about
how much better the preferred outcome. Instead of assigning numbers to the
outcomes, in the case of ordinal utility, it would be clearer if we just assigned
ranks, like instead of 1,2,3,4 and so on, we used 1st, 2nd, 3rd, 4th (and in
cases of indifference: for example, tied for 7th). In the cartoon figure about the
O RDINAL PREFERENCES Ordinal preference
rank outcomes: e.g. (x, y)
( x 0 , y0 )
(x00 , y00 ), without specifying how much (x, y)
is preferred to (x0 , y0 ) or (x0 , y0 ) is preferred
to (x00 , y00 ). The assignment of numerical
utilities representing ordinal preferences is
meaningful only to express the ordering:
u(x, y) > u(x0 , y0 ) implies only that the first
bundle is preferred to the second but not by
how much.
Planting in Palanpur game (Figure 1.2), we listed the four possible outcomes
as "Best, Good, Bad" and "Worst": this is an example of ordinal utilities.
There is no way that we can say that the top ranked bundle is twice as good
as the second-ranked bundle or ten times as good as the tenth-ranked bundle.
Nor could we add up the ranks, saying, for example, that getting your second
ranked bundle and your third ranked bundle with equal probability is as good
as getting your first and fourth ranked bundle with equal probability. None of
these statements make any sense. This is why when dealing with decisions
involving risk, we used a cardinal measure.
So, when we introduced the Palanpur farmers’ uncertainty about when the
C ARDINAL PREFERENCE A cardinal utility
function assigns a number to each outcome, with the property that the ratio of
the numbers assigned to alternative bundles expresses the relative degree of the
preference for the alternative bundles. For
example, with a cardinal utility function,
u(x, y) = 10u(x0 , y0 ) = 5u(x00 , y00 ) means
that (x, y) is preferred ten times as much as
(x0 , y0 ) which is preferred five times as much
as (x00 , y00 ), and that (x, y) is preferred fifty
times as much as (x00 , y00 ).
other farmer or farmers would plant their crops, we needed to think about expected payoffs, which requires adding up the values that each farmer attaches
to an outcome. Because you cannot add up ordinal measures, we gave the
payoffs numeric values (the numbers in the payoff matrix) representing cardinal utility.
For some questions in economics the ordinal – better or worse – meaning
of utility is all we need to understand and predict the actions that people
will take. But in many situations, those involving risk and uncertainty, as we
have just seen, or in evaluating the effects of differing rules of the game –
policies to ensure competition in markets or concerning fairness, for example
M - C H E C K For simplicity, we generally
restrict our analysis to outcomes that can be
described in terms of two variables x and y,
though it is straightforward to generalize this
model to outcomes described by more than
two variables. The actor therefore makes
choices among "bundles" that combine
different amounts of x and y.
– addressed in Section 3.13, the cardinal measure is required.
Checkpoint 3.2: Utility and payoffs
Give examples of preferences that might lead people to act in ways that they
would regret.
3.3 Indifference curves: Graphing preferences
Indifference curves are a useful way to visualize a person’s preferences.
Let’s illustrate the concept of an indifference curve by Annette, who is choosing among differing amounts of kilograms of coffee (x) and gigabytes of data
(y).
I NDIFFERENCE CURVE The points making up
an individual’s indifference curve are bundles
– indicated by (x, y), (x0 , y0 ) and so on –
among which the person is indifferent, so
that u(x, y) = u(x0 , y0 ) and so on.This means
that all of the bundles indicated by points
making up an indifference curve are equally
valued by the person.
124
MICROECONOMICS
- DRAFT
10
10
9
a
8
Gigabytes of data, y
Gigabytes of data, y
9
7
6
5
b
4
3
c
2
1
a
8
uA > 4
Better than uA4
7
6
5
b
4
3
uA < 4
Worse than uA4
2
c
uA4 = 4
1
0
0
0
1
2
3
4
5
6
7
8
9
10
0
1
2
Kilograms of coffee, x
(a) Consumption Bundles
Every point given by the coordinates (x, y) in Figure 3.4a is a pair of the
quantities of the two goods, called a bundle. Points a, b, and c therefore
represent three bundles of differing amounts of coffee and data. Suppose that
Annette ranks the points a, b, and c equally – she is indifferent among the
three bundles – then these three points lie on the same indifference curve,
as shown in Figure 3.4 b. Her indifference curve represents the combinations
of bundles among which she is indifferent. This means that for either bundle
a – 8 gb of data and 2 kg of coffee – or bundle b – 4 gb of data and 4 kg of
3
4
5
6
7
8
9
10
Kilograms of coffee, x
(b) An indifference curve
Figure 3.4: One of Annette’s indifference curves:
coffee and data. The dark green indifference
curve uA1 represents all the combinations of x and
y that provide Annette (A) with the same level of
utility, 4. The blue area above and to the right of
Annette’s indifference curve shows combinations
of the amounts of coffee and data that provide
her with utility greater than 4. The light green area
beneath her indifference curve shows the bundles
of x and y that she values at less than 4. She would
therefore rather choose a combination of x and y on
the indifference curve shown than any point to the
left or below it.
coffee– or bundle c – 2gb of data and 8 kg of coffee, u(2, 8) = u(4, 4) =
u(8, 2) = 4.
Figure 3.4 b shows the indifference curve made up of all bundles for which
Annette’s utility is equal to 4. Her indifference curve is labeled by a u with a
subscript which represents the level of utility that is the same for all points on
that indifference curve. Annette prefers to consume more of both data and
coffee, so she would like to be anywhere in the blue-shaded area where her
utility would be greater than 4). She would rather not consume less of both
data and coffee, so she would not like to be down to the area shaded in green
where her utility would be less than 4.
The single indifference curve shown in Figure 3.4 b divides the space of all
possible bundles of x and y into three categories: bundles that are respectively better or worse than any of the bundles making up u4 and bundles that
are equally valued with a utility of 4.
To understand a decision-maker’s choice we proceed in steps:
• Step 1: In this and the next section we use many such indifference curves
to evaluate all of the possible outcome of the decision.
• Step 2: In Section 3.6 we then limit the decision maker’s choices to those
B UNDLE A bundle is a particular allocation,
in the case of two goods given by (x, y). A
bundle that results from the choice of one or
more decision makers is call an outcome
bundle.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
Gigabytes of data, y
10
9
e
8
a
7
6
d
5
uA5
b
4
125
Figure 3.5: An indifference map for kilograms of
coffee, x, and gigabytes of data, y. The quantity
of good x is on the horizontal axis and the quantity
of good y is on the vertical axis. Three indifference
curves are shown: uA3 , uA4 , and uA5 , where the rank
of the utilities is uA5 > uA4 > uA3 . The constant level
of utility for uA4 = 4. Points a, b, and c all lie on
u2 and give Annette the same utility of 4. Point d
would give Annette lower utility and point e would
give Annette higher utility (because every bundle
is associated with some utility number, we could
draw indifference curves through those points, and
through any point in the figure).
3
c
2
uA4
1
uA3
0
0
1
2
3
4
5
6
7
8
9
10
Kilograms of coffee, x
that are feasible for (that is, choices that are actually open for the decision
maker to take).
• Step 3: Finally, use the evaluations in Step 1 to rank all of the feasible
outcomes, showing us the one the decision-maker ranks the highest.
To take Step 1, an individual’s utility function allows us to rank all of the outcome bundles – all combinations of x and y – that the decision maker considers. We do this using what is called a set of indifference curves (also termed
an indifference map) as shown in Figure 3.5.
Figure 3.5 shows three indifference curves, u3 , u4 , and u5 , part of Annette’s
indifference map. Annette prefers more of both goods – that’s why they are
called "goods." Therefore, indifference curves to the upper right, like u5 , are
higher, (corresponding to the blue-shaded area in Figure 3.4). Indifference
curves representing less preferred combinations, like u1 are to the lower
left (corresponding to the green-shaded area in Figure 3.4). Of the three
indifference curves plotted on the indifference map of Figure 3.5, uA
1 provides
Annette with her lowest utility, whereas uA
3 provides Annette with her highest
utility. A different person, one who valued coffee more than Annette would
have a different indifference map.
If you think of her indifference curves as a kind of contour map, Annette can
be pictured standing somewhere on a mountain wanting to get to the top.
She might, for example be in the lower left corner of the contour map of a hill
shown in Figure 3.6 wanting to reach the 800 meter top of the hill.
Her utility is the altitude where she is standing, say, at a point on the 720
meters above sea level contour. Her indifference curves are the numbered
contour lines on a map of the mountain she is climbing, each indicating loca-
I NDIFFERENCE MAP An indifference map
is a set of indifference curves selected
so as to illustrate some concept or result.
For example to compare two bundles or
to identify an outcome bundle that is the
outcome of the decision-making process.
126
MICROECONOMICS
- DRAFT
tions on the mountain the same height above sea level.
A map, as the quotation at the beginning of this chapter reminds us, is a representation of territory. The territory represented by Annette’s indifference
map is her evaluation all possible outcomes she might experience. An indifference curve runs through every point in the (x, y) plane, but just like maps
that could not possibly show every contour line, we can plot only a selected
number of them in any case.
Annette wants to climb as high as she can up the utility-mountain as possible,
given whatever limitations she faces, including her own physical capacities
and possibly impassible cliffs blocking her way. As Annette advances up the
mountain, she crosses contour lines, moving from lower to higher indifference
curves. She is engaging in a constrained optimization problem.
Checkpoint 3.3: Maps, Points and Bundles
Sketch your own version of the indifference map in Figure 3.5. Add two new
points to your graph:
a. A bundle, labeled f, where Annette holds the same amount of y as she does
at point b, but Annette prefers bundle b to f.
b. A bundle, labeled g, where Annette holds the same amount of y as she
does at bundle b, but which Annette prefers to bundle b.
c. Having manipulated the graphs and thought through the ideas of indifference
curves, explain why the following is true: Consistency of preferences implies
that indifference curves cannot cross.
3.4 Marginal utility and the marginal rate of substitution
Indifference maps are used to summarize the values that an individual places
on differing bundles of goods. But goods need not be things like Annette’s
coffee or data. Goods can be anything a person values, such as free time.
(Indifference curves, as we will show later in the chapter, can also summarize
the preferences people have about "bads" such as environmental degradation,
that, unlike goods, are things that people would prefer to avoid).
To see this, we will move from the choice about coffee and data, and think
instead about a new person, Keiko (KAY-i-ko), who is a student making a
choice about the use of her time. As Keiko progresses through her studies (no
doubt fueled by coffee and using data), she has two important priorities, which
she thinks of as "Living" and "Learning."
• Learning comprises all the aspects of her life as a student that contribute to
her goals of becoming an educated person and becoming qualified for an
interesting career.
Figure 3.6: A contour map of a hill showing
altitudes. Indifference curves are similar to contour
lines, which are composed of all the points in the
landscape which are at the same altitude. The
lower left quarter of the contour map resembles the
indifference map in Figure 3.5.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
127
• Living comprises everything else, including keeping up with friends, meeting new people, and taking care of herself.
As there are only so many hours in a day, and because Learning takes time,
Keiko faces what is called a trade-off between Learning and Living, the more
she has of one the less she will have of the other. So she is facing another
constrained optimization problem.
We explain in this chapter’s last but one section that constrained maximization
is not a description of the the mental and emotional processes by which we
adopt one course of action over another. It is a research strategy that we use
T RADE - OFF A trade-off is a situation in
which having more of something desired ( a
"good") requires having less of some other
"good" or more of something that the actor
would like to have less of (a "bad").
to understand what people do, not how they come to do it. But to illustrate the
method we will suppose the Keiko consciously maximizes her utility function
subject to her only-24-hours-in-the-day constraint, by comparing the utility
associated with each of the combinations of Learning and Living that are open
to her. (OK only a student in economics would actually do this!)
Keiko is a systematic and quantitatively oriented person, and decides to measure her Learning quantitatively with a number. In calculating her Learning,
she takes account of feedback from her teachers, such as grades (marks), but
also evaluates this feedback in terms of her own estimation of how much she
has learned, such as how much her study is improving her writing skills and
general understanding.
Keiko measures her the amount of Living by the hours she can spend not
studying, x, and the amount of her Learning by her personal rating, y.
Key to how the preferences, beliefs and constraints approach works is the fact
that for most of the things that we may value, if we have little of it, we highly
value having more of it, but the more of the thing we have, the less valuable
D IMINISHING M ARGINAL U TILITY What is
sometimes called the "Law of diminishing
marginal utility" holds that the marginal utility
of any thing that we value is less the more of
it that we have.
will be the next additional unit that we could have. This is called diminishing
marginal utility, where the new idea here is "marginal."
M-Note 3.1: The meaning of marginal
The change in the value of a function – like utility, u(x, y) – when just one argument of the
function x or y changes is a basic concept in calculus. The partial derivative of the function
with respect to an argument – that is either ux (x, y) or uy (x, y) – is an approximation of the
effect of a small change in the argument on the value of the function, holding constant the
other arguments. If the decision-maker increases her consumption of x by a small amount
Dx, then her utility is u(x + Dx, y) ⇡ u(x, y) + ux (x, y)Dx.
The marginal effect on utility of a change in some element in the bundle a person is consuming (x) is calculated as the size of the change in u relative to the size of some small
change in x with no other changes in the bundle. This is the size of the effect (Du) divided
by the size of the cause (Dx ), so ux (x, y) = Du
Dx where Dx is small. Conventionally this is
expressed as the effect on u of a one unit change in x.
If the marginal utility of any thing that we value positively is less, the more of it that we
have – diminishing marginal utility – then this means that:
F AC T C H E C K Diminishing marginal utility
in economics is often based on the psychological principle of satiation of wants, which
states that satisfying our wants is pleasurable, that our wants (for example hunger)
are limited, when the resources allowing
satisfaction of wants are limited we satisfy
our most urgent wants first, and that the
more satisfied is the want (by eating) the less
pleasure do we derive from further satisfying
the want.
128
MICROECONOMICS
- DRAFT
6
1
Marginal utility of Living, ux
g
Utility, u(x, y)
5
i
4
3
f
2
1
0
f
i
g
0
0
2
4
6
8
Living, x
10
12
14
16
(a) Utility of Living holding Learning constant
0
2
4
6
8
10
12
14
16
Living, x
(b) Marginal utility of living holding Learning constant at 3
• the first partial derivative of the utility function with respect to good x is positive, ux > 0,
(because more x is better than less) and
• the second partial derivative of the utility function with respect to x, uxx < 0 (because
as x increases, utility is increasing (ux > 0,) but at a diminishing rate, therefore giving
us diminishing marginal utility).
Diminishing marginal utility
A change of one variable – like Keiko’s Living – by one very small unit while
holding constant everything else, including her Learning, is a marginal
change, meaning the the change is very small and in only one variable. The
Figure 3.7: Diminishing marginal utility. In panel
a, utility is an increasing and concave function
of Living, meaning that the curve it is positively
sloped, but with a decreasing slope for higher levels
of Living. The slope of the curve is the marginal
utility, and this is shown in panel b. The points the
in panel a, correspond to the same points in panel
b. For example, the height of point f in panel a
shows the level of utility when Keiko experiences
2 just hours of Living and the slope of a tangent
to the curve at that point is the marginal utility of
Living that point. The height of point f in panel b
shows the value of that slope, that is the marginal
marginal utility of increased Living when Keiko
experiences 2 hours of Living.
change in utility corresponding to a marginal change in x or y is called the
marginal utility of x or y. Keiko’s marginal utility of Living which we denote as
ux , like her utility itself, depends on how much Living and Learning she is currently experiencing. So we write ux as a function of x and y: ux (x, y).
Similarly, Keiko’s marginal utility of Learning, uy (x, y) (or using the alternative
notation
Du(x,y)
Dy ,
is how much her utility changes as she changes her Learning
(y) by one unit, holding constant the amount of Living she does (x).
To understand marginal utility, let us compare points f, i and g in Figure 3.8.
By comparing these three points we can see how the marginal utility of Living
(x) changes while holding Learning (y) constant. Keiko’s amount of Learning
(y) is the same as she compares f, to i to g and the increase in her Living reA
A
sults in Keiko’s utility increasing from uA
1 to u3 to u3 . As her Living increases,
however, each additional increase in x is associated with a smaller increase in
utility, as reflected by her indifference curves getting flatter as she increases
the amount of Living (that is, moving to the right along a horizontal line in the
figure). This shows that the marginal utility of Living is decreasing as she gets
more Living (increased x).
Figure 3.7 shows just a slice of Keiko’s preferences, namely how they vary
M-CHECK
We also use the symbol for partial
∂ u(x,y)
differentiation ∂ x to mean the marginal
utility of x. When it is not necessary to
be reminded of the other variables (held
constant) that the marginal utility depends
on, we eliminate the (x, y) and just use ∂∂ ux or
ux .
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
Learning, y
Figure 3.8: An indifference map portraying the
choices between Living (x) and Learning y. The
negative of the slope of the indifference curve is
the marginal rate of substitution of Learning (y)
for Living (x), mrs(x, y), capturing the trade-offs of
Keiko’s preferences for the two goods. At f, Keiko
has a high level of Learning (3) and little Living (2
hours) and she is willing to give up a lot of Learning
to get more Living (her slope is steep at point f,
therefore her marginal rate of substitution is large).
At h, Keiko has a low level of Learning (0.82) and a
lot of Living (14 hours) and she is willing to give up
very little Learning to get more Living (her slope is
relatively flat at point h, therefore her marginal rate
of substitution is small). Keiko has a Cobb-Douglas
utility function with u(x, y) = x0.3 y0.7 .
Higher
utility
4
g
i
f
3
uA3
2
uA2
Lower
utility
1
h
uA1
0
2
4
6
8
10
12
14
129
16
18
Living (hours), x
with the level of Living she experiences, when the level of Learning she experiences is fixed at y = 3. We can study the full range of her preferences
when the values of both goods varies by looking at her entire indifference
map.
The marginal rate of substitution
This is shown in figure 3.8 where points f, i and g correspond to the same
points in the previous figure.
At point f, Keiko spends 14 hours studying and attending classes and has 2
hours left over for Living, with the result of a lot of Learning and not so much
Living. At point h, Keiko spends 2 hours studying and attending classes and
has 14 hours left over for Living, but her Learning is lower than at point f.
Comparing between points f and h, we can see that Keiko sacrifices some of
one good to get more of the other. Comparing f to h, Keiko gives up Learning
to get more Living, but her utility remains the same.
If we apply the same reasoning to very small differences in the quantity of
the two goods, we can see that at point f the largest amount of Learning that
130
MICROECONOMICS
- DRAFT
Keiko would be willing to give up in order to get one more unit of Living is the
negative of the slope of her indifference curve at that point (0.64 at point f),
which is the marginal rate of substitution at that point or mrs(x, y).
The fact that the indifference curve is steep at that point means she would be
willing to give up a substantial amount of Learning to get a little more Living.
This is because – as is clear from the previous figure – at point f she has
an ample amount of Learning and not much Living, so her marginal utility of
Living is high.
Or, to put it a different way: the fact that at point f she has a lot of Learning
means that the marginal utility of the Learning that she would give up to get
some more Living is not very large. So the opportunity cost to Keiko of trading
some Learning or some Living is low.
The marginal rate of substitution is the maximum amount of y that Keiko can
give up to get a small unit more of x without lowering her utility. The marginal
rate of substitution is also the amount of y that Keiko would view as substitute
for losing a small unit of x. The marginal rate of substitution should be read as
"units of good y per unit of good x."
We show in M-Note 3.2 that the marginal rate of substitution is equal to the
ratio of the marginal utilities of the two goods:
mrs(x, y) =
ux (x, y)
uy (x, y)
(3.1)
This is true because the amount of y that compensates Keiko for a small loss
of x is the ratio of her marginal utility of x, which tells us how much she misses
the x she has lost, to the marginal utility of y, which tells us how much she
appreciates the compensating gain in y.
Equation 3.1 also tells us something more about Figure 3.8. We can use the
idea of marginal utilities to understand Keiko’s marginal rate of substitution at
f, i and g in Figure 3.8.
At point f on indifference curve uA
1:
• The marginal utility of x is high because Keiko has very little x.
• Conversely, the marginal utility of y is low because Keiko has a lot of y.
• Therefore, her mrs(x, y) = uux is large and as you have already seen she is
y
willing to give up a lot of Learning to get a bit more Living.
But at point g on uA
3 the opposite is true:
• The marginal utility of x relative to the marginal utility of y is lower because
Keiko has a lot of x than she had at point f.
M ARGINAL R ATE OF S UBSTITUTION The
marginal rate of substitution is the negative
of the slope of the indifference curve. It
is also the willingness to pay for a small
increase in the amount x expressed as how
much of y the person would be willing to give
up for this. This is sometimes called the offer
price.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
131
• Therefore, her mrs(x, y) = uux is smaller and she is willing to give up very
y
little Learning to get more Living. This is why the indifference curve is flatter
at point g than at point f
This shows that because of diminishing marginal utility, for a given amount of
Learning, the more hours of Living Keiko has, the less amount of Learning she
is willing to give up to get another unit of Living.
The same reasoning shows (and Figure 3.8 confirms) that if Keiko’s preferences exhibit diminishing marginal utility for both x and y, her marginal rate of
substitution of y for x declines starting at point f as we consider points on an
indifference curve having more x and less y.
M-Note 3.2: The mrs is the ratio of marginal utilities
To derive the marginal rate of substitution using calculus, we use the method of total differentiation (covered in the Mathematical Appendix). First of all, along an indifference curve
the amount of utility is a constant, u(x, y) = ū.
So, to find the slope of the indifference curve we ask what changes in the quantities of x
and y (one increasing the other decreasing) are consistent with u(x, y) not changing. This
is what total differentiation tells us. The reason is that when we totally differentiate the
utility function with respect to its arguments we express the change in Keiko’s utility as the
sum of the changes due to changes in her consumption of each good.
The total derivative of this equation is:
du = ux (x, y)dx + uy (x, y)dy = d ū = 0
Since her utility is constant on an indifference curve by definition, the change in her utility
is zero.
Recall, too, that the derivative of a constant like ū is 0.
We can now re-arrange equation 3.2 to find the mrs(x, y):
Subtract uy (x, y)dy from both sides
Divide by uy (x, y) and dx
uy (x, y)dy
=
mrs(x, y)
=
ux (x, y)dx
dy
ux (x, y)
=
(3.2)
dx
uy (x, y)
dy
dx is equal to the ratio of
ux (x,y)
the marginal utilities of the goods, u (x,y) . But the negative of the slope of the indifference
y
As a result, the negative of the slope of the indifference curve
curve is the marginal rate of substitution of y for x, so we have shown that the marginal
rate of substitution is the ratio of the marginal utilities.
The mrs has the dimensions of an amount of good y per unit of good x because the
marginal utility of y has the dimensions utility per unit y, and the marginal utility of x has
the dimensions utility per unit x.
The mrs and the willingness to pay
The marginal rate of substitution provides us with an essential piece of information. Imagine that Keiko had some bundle (x, y) and she were offered the
following exchange – trade away some of her y in order to get more x. The
mrs tells us the greatest amount of y that she would be willing to give up to get
M C H E C K When considering two goods
– things that people value positively, like
data and coffee, or living and learning – the
indifference curves are downward sloping.
That is, they have a negative slope. The
negative of the slope of an indifference curve
is just its slope with the sign changed.
132
MICROECONOMICS
- DRAFT
one more unit of x in such a trade. This why we call the mrs the willingness
to pay y to get more x.
Why does the mrs tell us her maximum willingness to pay? The answer is that
if in return for another unit of x she gave up an amount of y equal to the mrs
she would be moving from one point on her indifference curve to another point
on the same indifference curve. This is how we constructed the mrs. So she
would be no better off after the trade than before.
She would happily pay less than the mrs to get one more unit of x because
this would increase her utility (put her on a higher indifference curve). But she
would not pay more. This is why we call the mrs the maximum willingness to
pay.
Before going on to the constraints facing Keiko we will now show how what
you have learned so far can be used with an explicit mathematical function.
3.5 Application: Homo economicus with Cobb-Douglas utility
In Chapter 2, we saw that people may be some combination of preferences
including self-regarding altruistic, fair-minded, reciprocal, spiteful, and so on.
Representing these preferences mathematically requires knowledge of what
Keiko values including:
• How important to her are Learning and Living?
• Is her own Living and Learning all she cares about, or does she value other
people’s Living and Learning.
In this section we study the preferences of a self-regarding Keiko: she does
not care about the Living and Learning of others. We use what is called a
Cobb-Douglas function utility function to illustrate how we can model the
difference it makes what value she places on the two elements in her choice
bundle.
Here is a Cobb-Douglas utility function:
u(x, y) = xa y(1
a)
(3.3)
The size of a , which is a positive number less than 1, is a kind of baseline
measure of how much the individual values x independently of how much x
and y she has. In M-Note 3.4 we show that if Annette is consuming the same
amount of x and y then the maximum number of units of y that she would be
willing to pay for one unit of x is 1 aa . So if a = 0.4 she her willingness to
pay for a unit of x would be 0.4
0.6 or two-thirds of a unit of y. We also show in
Chapter 7 that the fraction of a utility-maximizing consumer’s budget that will
be spent on good x is a . The fraction spent on y will be 1
a.
H I S TO RY The Cobb Douglas function
is named after the economist and later
U.S. Senator Paul Douglas and his then
Amherst College colleague, mathematician
and economist Charles Cobb, who jointly
came up with the function in 1928 for an
econometric study of the contributions of
labor and capital goods to output in the U.S.
economy.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
When a person’s preferences are described by a Cobb-Douglas utility function
then as long as the Keiko has some of each good, x > 0 and y > 0, the
following will be true:
• her utility uCD (x, y) > 0 is positive, and
• her utility increases as she consumes more of either good x or y, meaning
that the marginal utility of both goods is positive.
Because the marginal utilities for both goods is positive, Keiko will select
a bundle with more of each over one with less of either if both bundles are
available to her.
Here is an example of a Cobb-Douglas utility function where a consumer,
Annette from earlier, has a stronger preferences for y than for x because
a = 0.4 and (1
a ) = 0.6.
uCD (x, y) = x0.4 y0.6
(3.4)
Let’s assume that x is kilograms of coffee and y is gigabytes of data as we
did earlier. Because of the values of a and (1
a ), Annette has a stronger
preference for data than for coffee because a = 0.4 < 0.6 = (1 a ).
When a > 0 and (1
a ) > 0, Cobb-Douglas utility functions have the property
that a bundle must include some of both goods to be assigned a positive
utility, so we consider cases in which the person has x > 0, y > 0.
M-Note 3.3: Cobb-Douglas Diminishing Marginal Utility
How do we check that marginal utility is diminishing? Let us examine the marginal utility of
Living in the Cobb-Douglas utility function. For the moment, we keep the function general
with a :
u(x, y)
Utility Function
=
x a y1
a
(3.5)
To find the marginal utility of x we differentiate Equation 3.5 with respect to x:
Marginal utility of x
ux =
∂u
∂x
=
axa
1 1 a
y
(3.6)
For 0 < a < 1, the marginal utility of x is positive, that is ux > 0. Why? x and y are both
positive, as is the parameter a , as is the exponent 1 a . The exponent a 1 < 0, but this
simply means that x can be read as being in the denominator of the marginal utility. For
example, for a = 0.6, the marginal utility of x is:
ux
=
0.6
y0.4
y
=a
x0.4
x
(3.7)
You can see from Equation 3.7 that the larger is x the smaller will be the marginal utility
of x. To confirm that the marginal utility of x is diminishing, we need to differentiate the
marginal utility of x with respect to x. That is, we need to find the second derivative of the
2
utility function with respect to x, ∂∂ u2 x , that is, to partially differentiate Equation 3.6 with
respect to x.
133
134
MICROECONOMICS
- DRAFT
∂ u2
∂ 2x
Change in ux
Because 0 < a < 1, a
=
1 ) x (a
(a )(a
1 < 0. Therefore, a (a
2) (1 a )
y
<0
1) < 0. Therefore, the rate of change
of the marginal utility with respect to x is negative (marginal utility is diminishing), or what
is the same thing: utility increases at a decreasing rate as x increases.
Checkpoint 3.4: Positive utility for x and y
Consider Annette’s consumption of coffee and data as described by Equation
7.14.
a. Sketch a map of three indifference curves for Annette based on her utility
function.
b. Confirm that any bundle with positive consumption of coffee and data (x > 0
and y> 0), is assigned a positive utility.
c. How would you confirm whether consuming one more coffee or data increases Annette’s utility?
d. If either (both) coffee and data increase her utility, at what rate does Annette’s utility change for changes in her consumption of coffee or data?
(Hint: Mathematically, think through how you would find a "change in a
change").
M-Note 3.4: Cobb-Douglas Coffee & Data
We can derive the marginal rate of substitution for the general Cobb-Douglas utility function we defined earlier for coffee and data. Remember that along an indifference curve
utility is a constant, such as ū > 0. When we find the "change" of a constant like ū, that
change is zero, therefore du = 0 as in the set of equations below.
uCD (x, y)
=
xa y ( 1
du = ux (x, y)dx + uy (x, y)dy
=
0
a)
= ū
To find the marginal rate of substitution, we need to find the marginal utilities of x and y.
Consequently, we differentiate the utility function with respect to x to find ux , the marginal
utility of coffee, and with respect to y to find uy , the marginal utility of data.
ux
=
axa
uy
=
(1
1 (1 a )
y
a ) xa y ( 1
(3.8)
a) 1
(3.9)
We substitute the marginal utilities from equation 3.8 and 3.9 into the definition of marginal
rate of substitution, mrs(x, y) to find the formula for the marginal rate of substitution.
mrs(x, y) =
dy
dx
=
=
Factorize out x 1 and y 1
=
ux (x, y)
uy (x, y)
(1
axa 1 y(1
a ) xa y ( 1
a)
a) 1
axa x 1 y(1 a )
(1
a ) xa y ( 1
a)y 1
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
a
y(1
Remember that x 1 = 1x and y 1 1 = y and cancel the terms xxa and (1
y
mrs(x, y)
=
a
(1
y
a) x
a)
a)
135
:
(3.10)
For example, equation 3.10 shows that if Annette is consuming the same number of gigabytes of data and kilograms of coffee (say, 5 each) she will evaluate them at ratio (1 aa ) .
The preferences for each good (a and (1
a )) determines the ratio at which Annette is
willing to trade data for coffee, together with the amount of coffee and data she is actually
consuming. You can see that if Annette had a different level of current consumption of the
two goods, say, more x and less y her mrs would be lower.
y
Checkpoint 3.5: Diminishing mrs as x falls
Go back to Figure 3.5 and explain why the mrs is lower at point c than at point
b, and lower at point d than at point b. Can you say if the mrs is lower at point
a than at point d? Or at point e than at point a? Equation 3.10 may help you
answer this question.
3.6 The feasible set of actions
Keiko’s preferences and the resulting utility numbers she assigns to each
bundle are a reflection of what she wants to achieve, what her goals are. But
her preferences do not tell us what she can feasibly obtain. To understand the
bundles that are feasible for her, we need to know how she obtains Learning
from spending her time Studying. We suppose that Keiko sleeps 8 hours
every night and she is not considering changing that. Her choice is what she
will do with the 16 hours in the rest of the day .
The relationship between the time Keiko spends studying and the amount
of learning she achieves is given by a function that shows for the time (in
hours) spent Studying (h), how much Learning (y) results, y = f (h). This is
what economists call a production function – a mathematical description of
the relationship between the quantity of inputs devoted to production on the
one hand and the maximum quantity of output that the given amount of input
allows. Production functions are more often used to study things other than
success in coursework, that is outputs such as meals served, lines of code
written, or bushels of corn harvested.
Keiko’s production function is depicted in Figure 3.9 a. From it you can see
that to obtain Learning a Keiko must spend hours (h) Studying. Up to a maximum of 16 hours she can increase her learning by studying more. But starting
from studying just a few hours, doubling the amount of studying she does
does not double her Learning. We can see this by comparing points e’, i’ and
g’. Four hours of study (h = 4), gets Keiko y = 1.75 points of Learning, as
P RODUCTION F UNCTION A production
function is a mathematical description of the
relationship between the quantity of inputs
devoted to production on the one hand and
the maximum quantity of output that the
given amount of input allows.
136
MICROECONOMICS
- DRAFT
4
4
yfʹ = 3.75
yf = 3.75
gʹ
Studying to Learning
production function
y = f(h)
yiʹ = 3
Learning, y
Learning, y
i
yi = 3
iʹ
yeʹ = 1.75
g
eʹ
Infeasible
e
ye = 1.75
Feasible
set
1
1
Feasible
frontier
0
2
4
6
8
10
12
14
16
0
2
(a) The production of Learning by Studying
6
8
10
12
14
16
(b) The feasible frontier of Living and Learning
shown by point e’. But 8 hours gets her just 3 units of Learning, far from a proportional increase. This is because if she has just 4 hours she focuses on the
really important key points, while if she has 8 hours she gets into the details,
which add to her Learning, but not as much as the key ideas.
Keiko’s learning production function illustrates an important common economic phenomenon: diminishing marginal productivity. The marginal productivity of hours studying is the effect of a small increase in studying time on
the resulting Learning. As you can see from the fact that the production function in figure 3.9 is flatter for more hours of study, this marginal productivity of
Studying hours is diminishing.
This is similar to diminishing marginal utility. Just as the person satisfies her
most pressing needs if she has very limited expenditures, but can turn to frills
if she has more to spend, Keiko focuses on the essential points if her study
time is limited but can turn to the examples and further illustration if she has
more time to spend.
Because Keiko has two ways to use her time – Studying or Living – and her
waking hours are just 16 we know that:
Hours of Living = 16 Hours
4
Living (hours), x = 16 − h
Studying (hours), h
Hours of Studying
This makes it clear that
• She has just one decision to make not two: if she chooses hours of Studying, that also determines her Hours of Living.
Figure 3.9: Production of Learning by Studying
and the feasible frontier of Living and Learning.
Points e’, i’ and g’ on the production function show
combinations of hours of Study and the maximum
amount of Learning she could accomplish in that
time. Point i’ for example shows that if she studies
8 hours she could attain learning equal to 3 (she
could also attain less if she did spent the "studying"
time texting with friends.) The amount of Living
that she can have is her 16 hours minus the time
she spends learning, i.e. x = 16 h, as shown in
panel b. Panel b. shows the feasible frontier (dark
green curve), which is the border of the feasible
set (shaded in green). The feasible frontier is
just a flipped version of her production function.
Points beyond the feasible set (shaded in blue) are
infeasible or infeasible given the number of ours in
the day and her Learning production function. In
this figure the equation for the feasible frontier is
1 2
given by: y = 4 64
x .
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
137
• Because more time living means less time studying this means that the
opportunity cost of living more is some amount of learning less.
To see what this opportunity cost is see Figure 3.9 Panel b, showing the
feasible set of outcomes that Keiko might experience. The feasible frontier
shown there is the mirror image of the production function in the panel a. The
horizontal axis is no longer Studying hours but instead 16 minus Studying
hours, which is the amount of Living she can have for each level of Studying
she chooses.
At e, Keiko can Study for 4 hours, which means she is Living for 12 hours at e
and her Learning is 1.75. Or (point g) she could study for 12 hourw and have
learning of 3.75. All of the points like e, g, i and the rest of the feasible frontier
are choices that she could make.
The feasible set is the area bounded by the feasible frontier and the x and
y axes composed of all combinations of Living and Learning that she could
experience.
M-Note 3.5: A Living-Learning feasible frontier
A mathematical expression for the feasible frontier is:
Feasible frontier
y = ȳ
c(x)
(3.11)
The parameter ȳ is the maximum amount of y when x = 0, the y-intercept of the feasible
frontier: if c(0) = 0 then y = ȳ. The term c(x) is the cost of x, that is, how many units of y
(Learning) one must give up to get the value of x (Living) that she chooses.
Suppose Keiko’s feasible frontier between Living, x, and Learning, y, is described by the
relation:
y=4
1 2
x
64
The negative of the slope of the feasible frontier,
(3.12)
dy
dx
=
1
32 x.
Checkpoint 3.6
Redraw Figure 8.3 to show a new situation in which either:
a. Keiko discovers that by changing her diet she can get by perfectly well on 7
hours of sleep.
b. She transfers to a new university where it’s more difficult to get high grades.
3.7 The marginal rate of transformation and opportunity cost
Turning to Figure 3.9 we can also contrast two points on the feasible frontier
in, such as points a and b. At point a, Keiko spends 14 hours studying and
attending classes and has 2 hours left over for Living, with the result of a lot of
Learning and not so much Living. At point b, Keiko spends 2 hours studying
F EASIBLE FRONTIER The feasible frontier
is the border of the feasible set, showing for
any value of x the maximum value of y that
is feasible, meaning, that the decision-maker
can obtain.
138
MICROECONOMICS
Learning, y
4
- DRAFT
Figure 3.10: Utility maximization: Living and
Learning. Keiko’s feasible frontier for Living
and Learning is shown in green. Three of her
indifference curves are shown by uA1 , uA2 and uA3
in blue (uA3 > uA2 > uA1 ). She maximizes her utility
at the point on her feasible frontier on the highest
indifference curve, that is, at point b (where the
two are tangents). At point b, she maximizes her
utility where the marginal rate of substitution equals
the marginal rate of transformation by choosing to
spend 8 hours Living which gives her a subjective
Learning score of 3.
a
b
3
mrs(x, y) = mrt(x, y)
uA3
2
uA2
c
uA1
1
Feasible
frontier
0
2
4
6
8
10
12
14
16
18
Living (hours), x
and attending classes and has 14 hours left over for Living, but her Learning
is lower than at point a. The difference between the two points on the feasible
frontier illustrates another trade-off that is central to Keiko’s choice: more living
means less learning. And vice versa.
If we apply the same reasoning to very small differences of the two goods,
we can see that the opportunity cost in less learning that is required to
get more living is the negative of the slope of her feasible frontier at that
point, namely
mrt (x, y).
Dy
Dx .
This is termed the marginal rate of transformation or
The marginal rate of transformation is the smallest amount of y that Keiko
has to give up to get a small unit more of x.
The mrt is therefore Keiko’s
opportunity cost of x in terms of y or the minimum amount of y she has to
sacrifice in order to get a small unit of x. The interpretation of the mrt as
the opportunity cost of the x-good plays a major role in the reasoning in this
book.
Opp. cost of x
= - Slope of feasible frontier =
Dy
= mrt (x, y)
Dx
The marginal rate of transformation should be read as "units of good y per unit
H I S TO RY : F R E E L U N C H The 1975 collection
of essays by Nobel Laureate Milton Friedman titled There’s No Such Thing as a Free
Lunch: Essays on Public Policy popularized
the idea that there is an opportunity cost to
having more of anything that we value.2
M ARGINAL R ATE OF T RANSFORMATION
The marginal rate of transformation is the
negative of the slope of the feasible frontier.
It measures the sacrifice of the y-good
necessary in order to get more of the x-good.
It is therefore the opportunity cost of the x
good in terms of the y good.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
of good x." The term “transformation” is used because we think of a movement downwards and to the right along the feasible frontier as hypothetically
transforming (giving up) the y-good into (having more of) the x-good. There is
nothing actually being transformed.
You can also interpret the negative of the slope of the feasible frontier as how
much Learning you can get by giving up one unit of Living.
The opportunity cost of the x-good in therms of the y-good differs depending
how much of each good Keiko has. If her feasible frontier exhibits an increasing marginal rate of transformation as is shown in the figure, then her marginal
rate of transformation of y for x increases as she moves along the feasible
frontier toward having more x and less y. In this case, Keiko has to sacrifice
more y for x the more x and the less y she has.
As Figure 3.9 showed, between the y-intercept and point g:
• She needs to give up relatively little Learning (0.25 points) to get four hours
of Living.
• Keiko’s feasible frontier is relatively flat.
• Her marginal rate of transformation is therefore small.
• Therefore the opportunity cost of Learning for Living is low.
Between point e and the x-intercept, however, the opposite is true:
• She must give up a large amount of Learning (1.75 points) to get an additional 4 hours of Living.
• Her slope is steeper.
• Her marginal rate of transformation is higher.
• The opportunity cost of Learning for Living is greater.
Keiko’s feasible frontier demonstrates an increasing marginal rate of transformation, which is to say increasing opportunity costs of Learning, moving from
the y-intercept down the curve (left to right) to the x-intercept. This occurs
because of diminishing marginal productivity of studying time in producing
learning.
M-Note 3.6: The Marginal Rate of Transformation
Using Figure 3.9, we saw that the marginal rate of transformation was the negative of the
slope of the feasible frontier. Suppose the feasible frontier is described by the equation:
y=4
mrt =
1 2
x
64
dy
1
= x
dx
32
139
140
MICROECONOMICS
- DRAFT
In this case the mrt increases with x, and exhibits the property of increasing marginal
rate of transformation, or increasing opportunity cost. If you substitute the values we used
earlier, for four hours, eight hours and 12 hours of living, we can evaluate the mrt (x, y) at
each point:
1
• mrt (x = 4) = 32
(4) = 18
1
• mrt (x = 8) = 32
(8) = 28 = 14
1
• mrt (x = 12) = 32
(12) = 38
As x increases, the marginal rate of transformation increases, illustrating the idea of
increasing opportunity cost.
3.8 Constrained utility maximization: The mrs = mrt rule
From the feasible frontier we know that when maximizing her utility, the limited
time in Keiko’s day creates a trade-off. By combining the insights of feasible
frontiers and indifference curves – as in Figure 3.10 – we can understand how
Keiko will manage this tradeoff
• Constraints: She can choose some point on or within her feasible frontier
given by her production function and the limits of her time.
• Preferences: From among the points in her feasible set, she will prefer the
outcome bundle with the highest utility, meaning on the highest indifference
curve.
To understand Keiko’s constrained utility-maximizing problem, we contrast
points a, b, and c in Figure 3.10. An outcome bundle (x, y) is constrained
utility-maximizing if there is no other point in the feasible set with a higher
utility.
Point a is on Keiko’s feasible frontier and lies on indifference curve uA
1 . But, a
is not constrained utility-maximizing because Keiko could increase her utility
by increasing her Living time and decreasing her Learning, by moving along
the feasible frontier to the southeast. By similar reasoning point c cannot be
the highest indifference curve she can reach.
Keiko’s constrained utility-maximizing point is b in Figure 3.10, the point on the
feasible frontier that is on the highest indifference curve. We label it b because
it is the point where Keiko does the best she can.
Figure 3.10 suggests a useful way to think about Keiko’s constrained utilitymaximization problem. In the figure, we see that the constrained utilitymaximizing bundle is the point where Keiko’s indifference curve is tangent to
her feasible frontier. This means the indifference curve and the feasible frontier have the same slope at the constrained utility-maximizing point.
The slopes of the indifference curve and the feasible frontier express tradeoffs between the two goods. This is the basis of what we call the mrs = mrt
C ONSTRAINED UTILITY- MAXIMIZATION An
outcome bundle (x, y) is constrained utilitymaximizing if there is no point in the feasible
set that is on a higher indifference curve.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
rule.
M-Note 3.7: Equating mrs to mrt to find the constrained maximum
Suppose Keiko’s utility for Living (x) and Learning (y) is described by a Cobb-Douglas
utility function with parameter a = 0.4 and (1 a ) = 0.6:
u(x, y) = x0.4 y0.6
We find her marginal rate of substitution by finding the marginal utilities and substituting
uA
them into the equation mrs(x, y) = uxA
y
uAx
mrs(x, y) =
0.4(xA )
=
uAy
=
0.6(x )
uAx
uAy
=
2y
3x
0.6
A 0.4
(yA )0.6
( yA )
0.4
(3.13)
Suppose her feasible frontier is described by the equation:
y=4
1 2
x
64
Keiko’s constrained maximum must be on her feasible frontier. We find her marginal rate
of transformation by differentiating y with respect to x:
dy
1
= mrt (x, y) = x
dx
32
To find Keiko’s constrained maximum, we use the two expressions above for mrs and mrt ,
equating them to find a point on the feasible frontier consistent with the mrs = mrt rule:
2y
1
= x
3x
32
(3.14)
Then multiplying through by 32 x:
y=
3 2
x =4
64
1 2
x
64
x2 = 64
x=8
y=3
Keiko spends 8 hours on Living, studies 8 hours, and achieves a Learning level of 3.
Doing the best you can: The mrs = mrt rule
Summarizing the results so far, in Figure 3.10
1. The negative of the slope of the feasible frontier is the opportunity cost
of getting a unit more more of the x good, in terms of the amount of the y
good forgone.
2. The negative of the slope of an indifference curve is a measure of the
person s willingness to pay for a little more of the x good in terms of how
much of the y-good she would be willing to give up to get an additional unit
of the x good.
Using these two statements we can see why point a in in Figure 3.10 could
141
142
MICROECONOMICS
- DRAFT
not be the utility maximizing outcome bundle. The indifference curve is
steeper than the feasible frontier, so the value of getting more living exceeds
the associated opportunity cost (2 above is grater than 1) So she could do
better by giving up some Learning in favor of more Living.
REMINDER:
The opposite is true at point c: the feasible frontier is steeper than the indif-
• mrs, the marginal rate of substitution, is
the negative of the slope of an indifference curve.
ference curve, so the opportunity cost of having more Living falls short of the
value of an additional unit of Living. So she definitely would not want to give
up more Learning to get more Living.
mrs
AND
mrt
• mrt , the marginal rate of transformation,
is the negative of the slope of the feasible
frontier.
In fact, it means the opposite. By giving up a unit of Living she would get
a substantial increase in Learning (that is what the steep feasible frontier
means). Giving up a unit of Living could be compensated by a modest increase in Learning (that is what the flatter indifference curve means). So
the benefits of giving up some Living in return for more Learning outweigh
the cost.
So any point like a and c where the feasible frontier and the indif-
ference curve intersect cannot be the constrained utility maximizing output
bundle. This gives us the mrs
mrt rule: The the utility maximizing output
bundle is a point where
Slope of feasible frontier
= Slope of indifference curve
which requires that:
Marginal rate of transformation = mrt
= mrs = Marginal rate of substitution
Or, what is the same thing
Opportunity cost of x
= Willingness to pay for x
The rule expresses a simple and true idea: if the opportunity cost of something is less than your willingness to pay you should choose more of it (if you
can) and if the opportunity cost is greater than your willingness to pay, you
should choose less of it (if you can). But there are cases in which the utilitymaximizing outcome bundle is not a tangency of the feasible frontier and an
indifference curve:
• It may be that an difference curve is steeper than the feasible frontier, but
there is no way to get more of the x good. In this case the slope of feasible
frontier does not measure the opportunity cost of getting more of the xgood; that is impossible (its cost is infinite). The utility maximizing outcome
bundle at point b in Panel b of Figure 3.14 an example of a case – called a
corner solution – there the mrt = mrs rule does not work.
• We show in M-Note XX that there are conditions under which a bundle
such that mrt = mrs can also be a minimum not a maximum. We provide
an example of this in Chapter 6.
T HE mrs = mrt RULE In many of the models
that we consider in the remainder of this
book, the constrained utility-maximizing
outcome is a point on the feasible frontier
at which an indifference curve representing
the trade offs between the decision maker’s
objectives is tangent to the feasible frontier
representing the opportunity costs of having
more of one good in terms of the amount of
the other good foregone. This is the point
where the marginal rate of substitution is
equal to the marginal rate of transformation.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
Checkpoint 3.7: Changes in Keiko’s preferences
Find Keiko’s constrained utility-maximizing level of Living and Learning when
a = 0.6 and (1
a ) = 0.4, so that she values Living more than Learning.
M-Note 3.8: When the mrs = mrt rule fails
The rule can fail to identify the constrained utility maximum under two conditions: when
the maximum is a corner solution (so the rule is not satisfied) and when the rule is satisfied at a minimum rather than a maximum. Positing a case with diminishing opportunity
cost of obtaining one good in terms of the other good foregone will illustrate both cases
Setup. Assume that a person’s utility varies with the amount of goods x and y:
u(x, y)
x+y
=
and the feacible amount of good y is a function of good x:
y(x)
=
(1
x)2
(3.15)
The rule may select a minimum, not a maximum.The marginal rate of substitution and
marginal rate of transformation are:
ux
uy
dy
mrt (x, y) =
dx
Equyating the mrs and mrt
2(1 x) =
x⇤
=
y⇤
=
mrs(x, y) =
Using Equation 3.15
=
1
=
2(1
=
1
1
2
1
4
x)
Note that using x⇤ and y⇤, the utility is u = 34 . Alternatively, we could set (x, y) = (1, 0),
or (x, y) = (0, 1): both allocations are in the feasible set. In both cases, u = 1, which is
higher than the one that we have reached using the condition mrs = mrt .
The condition mrs = mrt will not give the utility maximum if the second order condition is
violated: the second derivative of the utility function with respect to the variables must be
negative. Let’s calculate it, replacing Equation 3.15 into the utility function:
u
d
dx
✓
du
= ux
dx
◆
du
= uxx
dx
=
x + (1
=
1
=
2>0
2(1
x)2
x)
The utility maximum may be a corner solution. In the example the utility maximums at
both x = 1 and y = 1 are corner solutions (only one of the goods is consumed.)
The rule may be inapplicable. Where either the indifference curves or the feasible frontier are not smooth but instead are kinked (are not differentiable), the derivatives on which
the mrs and mrt are based will not exist at some points.
Trade-offs between goods and bads
In many situations it is easier to understand decisions in terms of a trade-off
between a good and a bad rather than a trade-off between two goods. Recall
that a bad is something that you would prefer to have less of, such as working
143
144
MICROECONOMICS
- DRAFT
uA3
Higher
utility
uA2
3
b
2
mrs(x, y) = mrt(x, y)
3
Feasible
frontier
b
2
a
Lower
utility
1
2
uA1
c
a
0
uA2
4
c
Learning, y
Learning, y
4
uA3
uA1
4
6
8
10
12
14
1
16
18
0
2
Studying (hours = 16 − Living), h
4
(a) Indifference curves of Learning and Studying
8
10
12
14
16
18
(b) The utility-maximizing choices with bads
harder than is comfortable or safe.
For example, Keiko might think of her decision in terms of a trade-off between
her time studying time that she does not enjoy, h = 16
6
Studying (hours = 16 − Living), h
x, and her Learning,
y. The more time Living the better for Keiko, therefore x is a good. The more
time Studying the worse for Keiko, therefore h is a bad. But as before since
x = 16 h, choosing (h, y) to maximize utility, u(16 h, y) is the same thing
as choosing x to maximize utility u(x, y). These are just different ways of
posing the same problem.
Figure 3.11 shows Keiko’s indifference curves and feasible frontier plotted in
terms of Study time, h and Learning y. Her indifference curves slope upward
because an increase in Studying, h, lowers Keiko’s utility, and requires an
increase in Learning, y to compensate in order to stay at the same level of
utility. Utility increases as we move to the northwest and decreases as we
move to the southeast in this plot.
Similarly, Keiko’s feasible frontier slopes upward, because an increase in
Study time, h, leads to more Learning, y. This is her "learning production function" introduced earlier. So the slope of the feasible frontier is the marginal
Dy
productivity of studying time or Dh and this is also the marginal rate of transformation of Study time into Learning. (In this case "transformation" actually
describes the process underlying the feasible frontier).
As was the case for tradeoffs between two goods, a bundle in the feasible
set is the utility maximizing output bundle if there is no other feasible bundle
Figure 3.11: The mrs = mrt rule:. Keiko’s
problem of choosing (h, y) when h = 16 x =
time Studying, is a bad. Studying time, h is plotted
on the horizontal axis, and Keiko’s Learning, y is
plotted on the vertical axis. Keiko’s feasible frontier
is shown in green in the right-hand panel. Three
of her indifference curves are shown by u1 , u2 and
u3 in blue in both panels. The points a, b, and c
are the same as in Figure 3.10. Keiko maximizes
her utility at the point on her feasible frontier on
the highest indifference curve, that is, at point b,
choosing to spend 8 hours on Living and 8 hours
on Learning.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
145
with greater utility. And this is the bundle for which the mrs = mrt rule holds,
namely the point on the feasible frontier where the marginal rate of substitution equals the marginal rate of transformation (mrs(x, y) = mrt (x, y)).
M-Note 3.9: The marginal utility of the bad
The utility function for Studying (h) and Learning (y) is given by:
uA (h, y)
=
(16
h)0.4 y0.6
(3.16)
To find the marginal utility of the "bad," Studying, we need to partially differentiate Equation 3.16 with respect to h. Remember that when we partially differentiate we treat the
other variable as a constant, so the term y0.6 will simply remain where it is. We only have
to think about the h term.
∂ uA
= uAh
∂h
uAh
=
(0.4)( 1)(16
=
0.4 (16 h)
| {z } |
{z
<0
>0
h)0.4
1 0.6
y
0.6
y0.6 < 0
} |{z}
>0
The first term is negative whereas the second and third terms are positive. So the
marginal utility of hours of study is negative. We call such a utility a disutility and will
often talk about the disutility of work or the disutility of effort.
Checkpoint 3.8: Understanding goods and bads
Find Keiko’s constrained utility-maximizing level of Study time and Learning,
using the mrs = mrt rule.
3.9 The price-offer curve, willingness to pay, and demand
We often want to know how people respond to different options for exchange
in the form of prices. We may be interested in knowing, for each price at which
she can purchase any amount of the good she pleases, how much Keiko will
purchase, namely the utility maximizing amount. This is Keiko’s individual
demand curve.
Remember that in explaining Keiko’s indifference curves we asked what is the
maximum amount of Learning she would be willing to give up in exchange
for more Living. The answer is given by her maximum willingness to pay,
or what is the same thing her marginal rate of substitution of Learning for
Living.
We now ask almost the same question except that rather than giving up
Learning to get more Living, Keiko is now giving up money – that is paying
for a good according to its price.
For each offered price she faces another constrained utility-optimization problem. The demand curve is constructed by a series of hypotheetical constrained optimization problems, one for each possible offered price. Each
I NDIVIDUAL DEMAND CURVEAn individual demand curve (or demand function) indicates
for each price that might hypothetically be
offered at which a buyer can purchase any
amount that they please, the quantity that an
individual will purchase.
146
MICROECONOMICS
- DRAFT
y=m
y=m
Feasible
(within the budget)
Budget Constraint
Money left over, y
Money left over, y
a
Infeasible
(outside the budget)
yb
b
u3
u2
y=m−p⋅x
c
u1
Budget constraint, bc1
Kilograms of fish, x
x=
m
p
(a) The budget constraint
price defines a feasible set; its boundary, the feasible frontier, defines the bundles of goods Keiko has access to. For each each of these feasible sets there
is a bundle that maximizes her utility. This is a single point on her demand
curve.
Indifference curves tell us the utility number that Keiko assigns to each possible consumption bundle. Using this logic, her choice will be the point on the
feasible frontier with the greatest utility, which will be the bundle in the feasible
set that is on the highest indifference curve. This is a standard constrained
utility maximization problem.
xb
x=
Kilograms of fish, x
m
p
(b) Utility-maximizing choice
Figure 3.12: Budget constraint and utilitymaximizing choice for fish and money for
other goods. The budget set is shaded in green
and the budget constraint (feasible frontier) is
the dark green line on the border of the budget
set (feasible set). Consumption bundles (x, y) in
the budget set and on the budget constraint can
feasibly be obtained with the current budget (m)
at the price, p, for kilograms of fish, x, Outside the
budget constraint, in the shaded green area, the
bundles of x and y cannot feasibly be obtained
with the existing budget. Harriet maximizes her
utility subject to her budget constraint bc1 . She
maximizes her utility at b where her marginal rate
of substitution, mrs(x, y) = uux , equals her marginal
y
rate of transformation or the price ratio of x to y,
mrt (x, y) = p.
The budget constraint and feasible utility-maximizing choices
We shall use one particular kind of feasible frontier to think this through: the
budget constraint. The budget constraint defines an amount of money m
that a person has or has access to, through wealth and credit markets, which
constitutes their budget to spend on goods and services. People can use their
budget to spend on goods at prices that are given to them. Imagine that you
want to buy the fish that Alfredo and Bob were trying to catch in Chapter 5 at a
fish market. The price ( p) is measured in dollars per kilogram.
Figure 3.12 a. shows the budget constraint for Harriet, someone deciding on
how much fish to buy from Alfredo or Bob at price p. The budget set is shaded
in green and the budget constraint (feasible frontier) is the dark green line
on the border of the budget set (feasible set). Consumption bundles (x, y) in
the budget set and on the budget constraint can feasibly be obtained with the
current budget (m) at the price, p, for kilograms of fish, x, Outside the budget
constraint, in the shaded green area, the bundles of x and y cannot feasibly be
obtained with the existing budget.
B UDGET CONSTRAINT A person’s budget
constraint gives the bundles (x, y) that just
exhausts some given budget at a set of
market prices ( p) of the goods. The feasible
set includes all purchases bundles that
do not exhaust the budget, so the budget
constraint is the feasible frontier.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
147
We know how to find the utility-maximizing bundle for a given feasible frontier
– or the budget constraint – with given indifference curves: we apply the
mrs = mrt rule finding the bundle where the marginal rate of substitution
equals the marginal rate of transformation. We can combine these insights
and calculate what the consumer’s utility-maximizing bundle will be for every
potential price of the good given a fixed budget and when the other good, y, is
money for other goods.
Figure 3.12 b shows Harriet maximizing her utility subject to her budget constraint bc1 . To find her utility-maximizing choice, we must apply the mrs
mrt
rule to find where her marginal rate of substitution (her willingness to pay in
money for kilograms of fish) equals her marginal rate of transformation, here
the price for a kilogram of fish. At point a she consumes too little of x and too
much of y (her marginal utility of money for other goods (y) is much lower than
her marginal utility of kilograms of fish (x), or her mrs(x, y) is too high, and she
would be better off if she consumed less y and more x. Conversely, at c, she
consumes too little of y and too much of x (her marginal utility of x is much
lower than her marginal utility of y, or her mrs(x, y) is too low, and she would
be better off if she consumed less x and more y. She maximizes her utility at
b where her marginal rate of substitution, mrs(x, y) = uux , equals her marginal
y
rate of transformation or the price ratio of x to y, mrt (x, y) = p.
The demand curve: Utility-maximizing choices at difference prices
With every change in price, the consumer’s budget constraint will pivot.
The budget constraint will pivot upwards as a good’s price decreases, because a consumer can buy more of the good with the same budget. The opposite is true for price increases. As the price of a good increases, the same
budget buys less of the good, pivoting the budget constraint inward.
P RICE - OFFER C URVE The price-offer
curve shows every utility-maximizing consumption bundle at each price of good x. It
demonstrates the principle of demand by
connecting every point where a consumer’s
indifference curve is tangent to every possible budget constraint for a change in the
price of x at given income m. We will use the
price-offer curve in 4.
With every pivot of the budget constraint, at the utility-maximizing point, the
new budget constraint will be tangent to a new indifference curve which will be
either higher if the price of the good decreases or lower if the price of the good
increases.
Because we can calculate the utility-maximizing consumption bundle for
each possible price, we can find a curve that records every utility-maximizing
consumption bundle for each price, connecting up points a, b, and c in the left
panel of the figure. That curve is called the price-offer curve. Sometimes,
for individual consumers, it is called the price-consumption curve because it
indicates what the consumer will consume at different prices.
Figure 3.13 maps three different utility-maximizing consumption bundles
at three prices of x. With each price decrease, the budget constraint pivots
outward from p1 to p2 to p3 . With each change in the price of x, the utilitymaximizing bundle – the point at which the marginal rate of substitution is
M - C H E C K We can find the equation for the
price-offer curve by using the equation for
the budget line and combining it with the
equation for the marginal rate of substitution.
We do not derive it here as it is not required
to understand the intuition of the demand
curve.
148
MICROECONOMICS
- DRAFT
Money left over, y
Offer Curve
c
y3
b
y2
u3
p1 = 0.25
u2
a
y1
p2 = 0.5
u1
p3 = 1
Price per kilogram of fish, p
x1 = 6
p=2
p3 = 1
x2 = 9 x3 = 10.5
Demand
curve
aʹ
bʹ
p2 = 0.5
cʹ
p1 = 0.25
x1 = 6
x2 = 9 x3 = 10.5
Quantity of fish in kilograms, x
equal to the opportunity cost – changes. At p3 = 1, the bundle includes
x = 6, at p2 = 0.5, the bundle includes x = 9, and at p1 = 0.25, the bundle
includes x = 10.5. With each price change, there is a new bundle for both x
and y.
The different bundles suggest a price-quantity relationship between the quantity demanded of x and different prices of x. As the price of x decreases, the
quantity demanded increases. In fact, we can take each price-quantity combination and map a demand curve to it. In the lower panel of Figure 3.13,
we have taken each utility-maximizing consumption bundle from the different consumption bundles at each price and identified their coordinates on
price-quantity axes. The price-quantity combinations provide a downwardsloping demand curve where quantity demanded, x, decreases as its price, p,
Figure 3.13: Offer curve and demand curve for
fish: The price of x in the top panel is in terms
of the money Jane sacrifices to get more fish.
Similarly, in the lower panel the amount of money
Keiko must sacrifice to get more fish – the price
per unit of fish – determines Keiko’s quantity of fish
demanded along the demand curve. Points a, b,
and c in the top panel correspond to points a’, b’
and c’ in the lower panel.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
149
increases.
Measured horizontally from the vertical axis, it tells us the amount that can be
sold to the consumer at each particular price. Measured vertically from the
horizontal axis, it also tells us what is the consumer’s maximum willingness to
pay for each amount on the horizontal axis.
3.10
Social preferences and utility maximization
The preferences we have looked at so far have been entirely self regarding,
depicting a person who is concerned with their choices among bundles that
they alone will experience. But people often make choices where they are
not the only person affected, where what they choose can benefit or harm
E X A M P L E We demonstrate in Chapter 7
how to find the equation for the demand
curve and how to see that a reduction in
the price of fish has two effects. First, the
lower price leads the person to buy less
meat and more fish; this is the substitution
effect. Second, the lower price also allows
the person to buy more of everything if she
chooses (fish, meat or whatever); this is
called the income effect.
someone else. Consider the Dictator Game that we mentioned in Chapter 2.
In that game, a person, the Dictator has an endowment of money, y, that they
can choose to split between themselves and another person in any way they
choose.
Imagine that Annette is the Dictator and she is able to choose whether or
not to give to some amount to Ben. As a result, Annette must choose some
split of of her endowment z = p A + p B , where p A is the amount in dollars
that Annette keeps for herself and p B is the amount that she gives to Ben. As
a result, we can re-arrange the equation to find the equation to the feasible
frontier for y = 10 dollars:
Feasible Dictator Allocations
p B = 10
pA
(3.17)
Looking at Equation 3.17, we can see that the feasible frontier is a straight line
with a slope of
1. This tells us that the feasible frontier slopes downward.
Remember that the negative of the slope of the feasible frontier is the marginal
rate of transformation: so a player in the Dictator Game who wishes to give $1
to someone else has an opportunity cost of $1 for doing so. If Annette is like
Homo economicus, she is purely self-regarding. She sets p B = 0 and keeps
R E M I N D E R A game is a mathematical
representation of a strategic interaction,
which means one in which players recognize
that their payoffs depend on the actions
taken by other players. So the so-called
Dictator Game is not really a game at all,
because the Dictator’s payoffs do not depend
at all on anything that the other player does.
everything or herself and therefore z = p A = 10. What happens when the
Proposer is an altruist who believes in making an offer of more than zero to a
partner?
To see what happens in these cases, let us contrast two pairs of people:
M - C H E C K Two things to remember when
thinking about Equations 3.18 and 3.19.
• Annette (A) is paired with Ben (B). Annette makes choices about how much
money she gets and how much money Ben gets.
• Chen (C) is paired with Diane (D). Chen makes choices about how much
money he gets and how much money Diane gets.
To think about the choices that Annette and Chen make, let us consider two
different kinds of Cobb-Douglas utility functions that Annette and Chen might
• The exponents in the Cobb Douglas
utility function Equation 3.19 mean
that if they both had the same payoff,
then Chen would value increasing his
own payoffs more than he would value
increasing Diane’s.
• Any number raised to a zero exponent is
equal to 1, so because Annette does not
value Ben’s payoffs at all (the exponent
is zero) her utility is unaffected by the
amount that he gets (her utility is simply
how much she keeps for herself).
150
MICROECONOMICS
- DRAFT
11
uA1
uA3
uA2
11
9
9
8
8
D's payoff (dollars), πD
10
B's payoff (dollars), πB
10
Feasible
frontier
7
6
5
4
3
7
6
5
mrs = mrt
4
2
1
1
b
0
1
2
3
4
5
6
7
8
A's payoff (dollars), πA
9
10
bʹ
3
2
0
aʹ
uC
1
0
11
0
1
2
uA (p A , p B ) = (p A )1 (p B )0
(3.18)
uC (p C , p D ) = (p C )0.7 (p D )0.3
(3.19)
Equations 3.18 and 3.19 represent the utility functions of two different people.
Chen is other-regarding, he cares about Diane’s payoff as is indicated by the
positive exponent on her payoff in his utility function, though not as much as
he cares about his own (compare the two exponents). Annette is entirely selfregarding, placing a zero weight on Ben’s payoff and therefore her choice will
depend entirely on the outcome that she experiences.
We display indifference curves for Annette and Chen in Figure 3.14. The
indifference curves in panel a are unusual: they are vertical because the only
thing that Annette values is what is on the horizontal axis, namely, her payoff.
Using the mrs = mrt rule, we find the constrained utility-maximizing point
for each person where their highest indifference curve touches the feasible
frontier.
In this case, though, the feasible frontier is given by a straight line because it
represents a split of money. The maximum amount of money that Annette or
Chen can keep is $10 and they can offer splits in 1 cent increments between
themselves and their partners. The vertical intercept corresponds to the
instance in which they give all $10 to their partners. The horizontal intercept
corresponds to the instance in which they keep all $10 to themselves. Chen
has preferences such that he would like a 70%-30% split of the $10 (his
3
4
5
6
7
8
C's payoff (dollars), πC
9
10
11
(b) Chen is altruistic, offering a (7, 3) split
have.
Chen’s Utility Function
uC
2
cʹ
(a) Self-interested Annette offers a (10, 0) split
Annette’s Utility Function
uC
3
Figure 3.14: Utility maximization: Self-interested
offer vs. altruistic offer. Annette offers a split to
Ben of (10, 0), whereas Chen offers Diane a split
of (7, 3). Annette’s indifference curves are vertical
because she gives no weight in her utility function
to Ben getting any money ((1 a ) = 0), therefore
she gets $10 and Ben gets $0. Between Chen and
Diane, Chen gives some weight to Diane getting
money ((1 a ) = 0.3), therefore his indifference
curves are shaped like indifference curves we’ve
looked at previously and at his constrained utility
maximum Chen gets $7 and Diane gets $3. Notice
that if Chen gives any less or any more to Diane,
then he would be on a lower indifference curve,
such as at points b’ and a’ on uC1 .
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
a = 0.7) and his highest indifference curve is tangent to the feasible frontier at
a (7, 3) split shown by point b’ in Figure 3.14 b.
Annette has preferences such that she would like a 100%-0% split of the $10
(her a = 1, she places zero weight on Ben’s payoff) and her highest indifference curve touches the feasible frontier at b in Figure 3.14 a at a (10,0)
split (she keeps all the money). We can interpret the slope of her indifference
curves as her maximum willingness to pay in order to give Ben a small positive payoff, and the ask: how much of her own payoffs would she be willing
to give up to transfer a penny to Ben? The answer is that there is no amount,
however small, that would motivate her to do this.
But what allocation does she choose? She chooses the highest utility that
is within the feasible set. Her highest utility is where her vertical indifference
curve uA
2 touches her highest feasible allocation to herself of $10. She keeps
all the money. Her keeping all the money shouldn’t surprise us because she
gives no weight to Ben’s payoff. In mathematics, a solution like this is called
a corner solution. Notice that we couldn’t use our standard requirement for
finding the constrained utility maximum of mrs = mrt . mrs(p A , p B ) was
undefined because her indifference curves were vertical. But the principle
of constrained utility maximization, that Annette would find the point in the
feasible set with the highest utility, still applied to our problem and we found
the solution.
M-Note 3.10: The mrs for a self-regarding Dictator
Why are Annette’s indifference curves vertical in Figure 3.14? To answer this question,
we need to find her marginal rate of substitution. To find her mrs, we need the marginal
utilities of the two arguments of her utility function: p A and p B the money payoffs that
Annette and Ben respectively get.
Marginal utility to Annette of Annette’s payoff:
uAp A =
∂ uA
∂ pA
1 · ( p A ) (1
=
1)
(p B )0 = 1
Marginal utility to Annette of Ben’s payoff:
uAp B =
∂ uA
∂ pB
=
0 · ( p A ) 1 ( p B ) (0
1)
=0
Therefore Annette’s marginal rates of substitution is:
mrs(p A , p B )
=
=
up A
up B
1
= undefined
0
(3.20)
Now, the result of Equation 3.20 should not surprise us because the slope of a vertical
line is undefined. Annette’s indifference curves endlessly rise and have no run, so the
negative of an undefined number (the slope) remains an undefined number (the mrs).
Her indifference map therefore represents a range of vertical lines where the horizontal
intercepts correspond to the amount of money she keeps which is also the utility number
associated with the particular indifference curve.
151
152
MICROECONOMICS
- DRAFT
Now, we might ask ourselves, what is Annette’s utility at her constrained utility maximum?
Let’s substitute in the values we have for p A = 10 and p B = 0.
uA (p A , p B )
=
(p A )1 (p B )0
=
(10)1 (0)0 = 10
(3.21)
Annette has a utility that is equal to the amount of money she keeps for herself.
M-Note 3.11: An altruistic person splitting the pie
We will derive Chen’s decision about splitting the pie between him and Diane. Using
Equation 3.10, his marginal rate of substitution is (see his utility, Equation 3.19):
mrs(p C , p D )
7p D
3p C
=
Now, let’s assume that the size of the pie is z = 1, therefore, the feasible allocations are
represented by p D = 1 p C , so his mrt (the negative of the slope of the feasible frontier)
is
mrt =
dp D
dp C
=
1
Equating the mrs with the mrt , we can obtain how much Chen allocates to himself and to
Diane:
7p D
3p C
=
1
pD
=
3 C
p
7
=
1
=
1
=
0.7
=
0.3
3 C
p
7
10 C
p
7
)
pC
Using the feasible allocation set
and
pD
pC
That is why Chen offers Diane $3 of the total of $10 that she is able to allocate.
Checkpoint 3.9: Chen’s Choice and the mrs = mrt rule.
1. What is the marginal rate of transformation in this in the game described in
Figure Figure 3.14?
2. Why is the utility maximizing outcome bundle at point b in Panel b of Figure
3.14 an example of a case there the mrt = mrs rule does not work. How
does this case differ from the case shown in Panel b, where the rule does
work?
3. Use the value of Chen’s mrs at point c’ in Figure 3.14 Panel b along with the
value of the mrt to explain why for Chen the opportunity cost of giving more
money to Diane is less than his willingness to pay (give up his own payoffs)
so that Diane can have more.
3.11
Application: Environmental trade-offs
We think of environmental damage as something to be avoided, but stopping
or slowing the damage – or "abating" the damage in the language of environ-
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
153
mental science – is costly. Less damage means some combination of less
consumption, changing our consumption patterns to be less damaging to
the environment, or diverting our productive potential from producing goods
that we can now consume to discovering and installing new technologies.
We therefore face a trade-off between consuming goods and maintaining the
quality of the environment. How much of these opportunity costs of improved
environmental quality are we willing to pay?
The constrained utility maximization method we have developed provides a
way of posing and answering these questions using the preferences, beliefs,
and constraints approach.
Feasible combinations of conventional goods and environmental quality
The opportunity cost of environmental quality is consumption of other (conventional) goods such as food, clothing, shelter, and transportation, which we
must give up to secure a higher quality environment. There is a feasible frontier showing the combinations of environmental quality, x, and conventional
goods, y, that are possible for a society. The feasible frontier in the case of
environmental quality depends on the abatement technology, which represents how much consumption of conventional goods society has to give up to
achieve a given level of environmental quality.
Figure 3.15 shows a feasible frontier between conventional goods (y) and
environmental quality (x). We measure environmental quality on a numeric
scale from 0 (the environment that we would have if no abatement done) to
20 (the environment resulting if we were to divert to abatement uses all of
society’s resources above some minimum level of consumption). We measure
conventional consumption as billions of dollars.
The negative slope of the feasible frontier at any point is the marginal rate of
transformation of reduced environmental quality into increased conventional
consumption, or
Dy
Dx .
The steeper the frontier, the greater is the increase in
feasible consumption allowed by a reduction in environmental quality.
This is also the the opportunity cost of improved environmental quality. So a
flatter frontier means a lower opportunity cost of abatement.
To see this, starting at no abatement expenditures (y = ȳ), the opportunity
cost of improved environment is initially small (the frontier is nearly flat) and as
Annette implements more abatement, the cost more abatement increases as
the environmental quality increases. The shape of the feasible frontier reflects
an increasing marginal rate of transformation, or an increasing marginal
opportunity cost of environmental quality.
Put another way, if environmental quality is at its maximum at the intercept
of the feasible frontier with the horizontal axis, society could consume a lot
H I S TO RY In the middle of the twentieth
century, long before we worried about
climate change and its unfolding calamities,
Aldo Leopold, the American environmentalist
raised an economic question: "Like winds
and sunsets, wild things were taken for
granted until progress began to do away with
them. Now we face the question whether a
still higher ’standard of living’ is worth its cost
in things natural, wild and free." (Leopold
2020 [1949], p. xxi).
Leopold was articulating a trade-off
between, on the one hand, consuming goods
and services – Leopold’s higher "standard
of living" – and on the other, the costs of
environmental damage – the "cost in things
natural, wild and free."
154
MICROECONOMICS
- DRAFT
Figure 3.15: Trade-off between consumption of
conventional goods and environmental quality.
The constrained utility maximum is the point on the
feasible frontier on the highest indifference curve
u2 , shown as point b where the mrs = mrt rule
holds. The constrained maximum is at the point
where the feasible frontier is tangent to the highest
attainable indifference curve.
Policy−maker's
indifference curves
Goods in millions, y
y
Abatement
Cost
yb
b
u3
u2
u1
Initial
feasible
frontier
xb
x
Environmental quality, x
more conventional goods if it were willing to tolerate a small deterioration of
environmental quality (the frontier is steep where it intercepts the horizontal
axis). But the feasible increase in consumption of conventional goods allowed
by a reduction in environmental quality falls as the level of environmental
quality declines.
Checkpoint 3.10: The mrt of the environmental feasible frontier
a. Refer to Equation 3.22 and find the marginal rate of transformation of the
feasible frontier.
b. Practice sketching the feasible frontier and confirm the intercepts with the
horizontal and vertical axes as shown in Figure 3.15.
3.12
Application: Optimal abatement of environmental damages
How much abatement is the right level, taking account of both preferences for
conventional goods (consumption) and the quality of the environment along
with the opportunity costs in lost consumption?
A citizen chooses a level of abatement of environmental damages
To begin with the simplest case, think of just one citizen, Annette, who might
be representative of the attitudes of the whole society, trying to decide on the
M - C H E C K An example of the function
representing the feasible frontier is shown
below, where y is the goods available for
consumption, ȳ is the level of y that is
feasible when environmental quality is at its
minimum, and x is environmental quality.
y
=
100
1 2
x
2
(3.22)
This is the equation for the feasible frontier
is graphed in Figure 3.15.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
level of abatement that she would like to see implemented. She cares about
both the quality of the environment, x, and the amount of conventional goods
that will be available for people to consume
Annette’s utility function has the following form: u = u(x, y). Annette considers
what she would like to see her society do about the environment (x), taking
account of the effects on everyone. In other words, she is thinking from an
other regarding like an ideal policy-maker.
Annette’s indifference curves between environmental quality and conventional
goods are downward sloping because she regards both environmental quality
and conventional consumption as goods for which more is better. This means
the marginal utility of both y and x are positive (e.g. uy > 0 and ux > 0).
The negative slope of the indifference curves shown in Figure 3.15 at any
point is Annette’s marginal rate of substitution between more consumption of
goods and a better environment. Her marginal rate of substitution shows the
amount of goods she would be willing to give up for a small improvement in
the environment. As before, Annette’s indifference curves exhibit diminishing
marginal utility of both environmental quality and consumption.
An example of a utility function that Annette might have is the Cobb-Douglas
utility function:
u(x, y) = xa y(1
a)
= x0.4 y0.6
(3.23)
Figure 3.15 shows three indifference curves defined by equation 3.23: uA
3
is unattainable given the feasible frontier, uA
1 intersects the feasible frontier
twice, and uA
2 is tangent to the frontier at point (xb , yb ). Annette’s constrained
maximum allows her and her fellow citizens to consume 75 million units of
conventional goods and enjoy environmental quality of about 7 (see M-Note
3.12 for the worked solution). If she were able to implement relevant environmental and fiscal policies, this point is the best society can do in Annette’s
opinion.
What is the total opportunity cost in foregone conventional consumption of a
level of environmental quality of 7? The maximum feasible level of conventional consumption with no abatement is $100 billion. The difference between
the maximum feasible consumption of $100 billion and Annette’s preferred
choice of conventional consumption of $75 billion is the opportunity cost of
an environmental quality of 7. In our example, the abatement costs are equal
to $100 billion
$75 billion = $25 billion in conventional goods. A citizen
with Annette’s preferences thinks that the sacrifice of $25 billion consumption
goods is more than worth paying to have an environmental quality of 7 instead
of zero.
155
156
MICROECONOMICS
- DRAFT
New technologies, and conflicts of interest
If, with a mind to the future, some of the abatement costs are devoted to
research to improve abatement technologies, this would pivot the feasible
frontier outwards, as shown in Figure 3.16. Remember this is very similar
to the expansion of the feasible set shown in Figure Figure 8.3 when Keiko
adopted improved studying so as to reduce the opportunity cost in reduced
Living time associated with greater Learning.
As is shown in the figure, the shift of the feasible frontier would permit higher
environmental quality of x ⇡ 9.8 at the same level of consumption of $75 billion
at the new (xr , yr = yb ). But there would still be a trade-off: more conventional
goods would require less environmental quality, or more environmental quality
would require fewer conventional goods to stay on the feasible frontier.
We can also use Figure 3.15 to see why people often disagree about environmental policy.
• Preferences: peoples’ preferences for conventional goods and the environment may differ
• Beliefs: people may disagree about the opportunity costs or the benefits of
environmental quality
• Conflicts of interest: the costs and benefits of abatement fall on different
people; those whose jobs or profits depend on carbon based energy, for
example, stand to bear more of the costs of addressing climate change,
while regions likely to be particularly hard-hit like Africa bear a larger share
of the benefits.
M-Note 3.12: The trade-offs and opportunity costs of the environment
Let us work through the process that Annette the policy-maker would go through to identify
the combination of goods in billions of dollars with environmental quality.
First, let us calculate her marginal rate of substitution from her utility function,
uA (x, y) = (xA )0.4 (yA )0.6 . From earlier in the chapter, we know that the mrs(x, y) is the ratio of marginal utilities and we have already calculated this for a = 0.4 and (1 a ) = 0.6
in Equation 3.13 in M-Note 3.7.
mrs(x, y)
2y
3x
=
(3.24)
Annette’s feasible frontier, based on her beliefs and understanding of the existing science,
dy
is given by the equation y = 100 12 x2 , for which we can find her mrt (x, y) = dx :
)
dy
dx
dy
dx
x
=
=
x
(3.25)
We now set the mrt (x, y) given by Equation 3.25 equal to the mrs(x, y) given by Equation
F AC T C H E C K The pace of environment
friendly innovation is astounding. Have
a look at the reduction in costs of the
photovoltaic cells used in solar panels
dropping to one-onehundreth of there costs
in 1975 in Figure 8.3.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
3.24 and we isolate one of the variables, y:
Multiply through by 3x
Divide through by 2
x
=
3x2
=
y
=
2y
3x
2y
3 2
x
2
(3.26)
We can now substitute Equation 3.26 into the feasible frontier to find xb and yb :
3 2
x
2
2x2
=
100
1 2
x
2
=
100
x2
=
) xb
=
50
p
50 = 7.07
Having found xb , we can substitute it back into 3.26 to find yb :
y
y
b
=
100
=
100
=
75
1 p 2
( 50)
2
1
(50)
2
So, as a result of Annette’s policy-making utility function and feasible frontier, she would
choose a combination of environmental quality, x, of value 7.07 with consumption of good
and services of $75 billion. $75 billion is $25 billion less than the maximum consumption
of goods and services, ȳ = 100, so the cost of abatement is $25 billion.
First, people may differ in their preferences over conventional goods and the
environment. Another citizen, Brenda, may not worry as much as Annette
about the climate and problems that future generations will inherit because
she puts a lower weight on the welfare of others (in this case of future generations). Brenda would have different indifference curves. If she also had
Cobb-Douglas utility, she would have less strong preferences for the environment than Annette, with a lower a and higher (1
a ). Brenda would as a
result choose a constrained maximum with lower social spending on abatement.
A second reason for disagreement is that Brenda thinks that the costs of
abatement are much greater than Annette thinks they are.
If Brenda thinks that the actual costs of environmental quality are greater than
Annette does, Brenda would work with a different feasible frontier inside the
feasible frontier Annette considers. Brenda’s feasible frontier would be steeper
over its entire range, indicating that the opportunity cost of environmental
quality is higher for any level of environmental quality than on Annette’s feasible frontier. Brenda’s constrained utility-maximizing bundle would be different
from Annette’s.
But there is a third reason for disagreement, not having to do with the preferences or beliefs of the citizens. Some members of the society may benefit
from decisions that harm the environment. Brenda might be employed by or
157
158
MICROECONOMICS
- DRAFT
Figure 3.16: Trade-off between consumption of
conventional goods and environmental quality
with R&D. The choice between consumption of
conventional goods and environmental quality with
R&D pivoting the feasible frontier outwards leading
to a new point of tangency on a higher indifference
curve u3 .
Policy−maker's
indifference curves
Goods in millions, y
y
Abatement
Cost
yb
r
b
u3
u2
u1
Initial
feasible
frontier
xb
xr
Feasible
frontier
with R&D
x
Environmental quality, x
an owner of a firm producing fossil fuels, for example. They will bear more
than proportionally the costs of abatement and may prefer lesser levels of
abatement for that self-regarding reason.
The fact that a policy of CO2 emissions abatement would affect Brenda adversely while benefiting Annette brings us back to the how we understand the
term utility.
Checkpoint 3.11
a. Show that when Annette’s utility function is defined by Equation 3.23 and
the feasible frontier is defined by Equation 3.22 with ȳ = 100, that the utilitymaximizing consumption bundle is $75 billion with an environmental quality
of 7.07.
b. Show that when Annette’s utility function is defined by Equation 3.23 and
the feasible frontier is defined by Equation 3.22 with a technological im1 2
provement where y = 100
3 x that the constrained utility-maximizing
consumption bundle is (xr = 8.66, yr = $75 billion).
c. Show that when Annette’s utility function is defined by Equation 3.23 and the
feasible frontier is defined by equation 3.22 with a technological improve1 2
ment where y = 100
4 x , the constrained utility-maximizing consumption
bundle is (xr = 10, yr = $75 billion). Draw a new feasible frontier tangent
to a new indifference curve u4 with accurate intercepts reflecting the new
technology.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
3.13
159
Cardinal inter-personally comparable utility: Evaluating policies to reduce inequality
Most policy choices involve conflicts of interest like that between Annette and
Brenda about the abatement of environmental harms. Few policy choices are
entirely win-win. Most policies – whether they concern taxation, immigration,
health insurance, or the rate of inflation – result in benefits for some and
losses for others.
Ordinal and cardinal utility in policy evaluation
How do we then evaluate competing policies? Don’t think about this as a
question about what would be a good outcome for you if you were a participant in the society. Instead, try to take the position of what Adam Smith called
the Impartial Spectator who did not himself stand to gain or lose, but wanted
instead to consider the gains and losses to society.
One answer you might give is just to count those who prefer each policy and
select the most popular policy. All this requires is that people be able to rank
H I S TO RY Adam Smith in The Theory of
Moral Sentiments conceived of the impartial
spectator as follows, "We endeavour to
examine our own conduct as we imagine
any other fair and impartial spectator would
examine it. If, upon placing ourselves in his
situation, we thoroughly enter into all the
passions and motives which influenced it, we
approve of it, by sympathy with the approbation of this supposed equitable judge. If
otherwise, we enter into his disapprobation,
and condemn it."3
the policies in question as better, worse, or indifferent. We could in this case
treat utility as ordinal (that is, simply a ranking (or ordering) of outcomes).
Something like this might occur in a majority rule democratic political system,
especially if citizens could vote on policies as they do in many countries in
referendums asking citizens to vote for or against a particular policy.
But this way of evaluating policies might result in evaluating positively those
policies that confer minor gains to those in favor, and substantial losses to
those preferring another policy. This does not seem like a sensible rule.
An alternative is to weigh the amount of the gains to the beneficiaries of each
policy against the size of the costs incurred by those who would have done
better under some other policy. This kind of comparison requires that we know
not only which policies people prefer, but how much they prefer them.
To do this we treat utility as a cardinal measure for which utility is not just a
ordinal ranking, but instead a number indicating how well off the person is
under the option in question. Treating utility as cardinal allows us to say two
very different things:
1. for Annette, the outcome (x0 , y0 ) is twice as good as (x, y) because for
example uA (x0 , y0 ) = 2uA (x, y)
2. the sum of the Annette’s and Brenda’s utility is greater with outcome (x0 , y0 )
than with outcome (x, y) because uA (x0 , y0 ) + uB (x0 , y0 ) > uA (x, y) +
uB (x, y)
Both statements involve cardinal utilities, but they differ. The first statement
compares how much Annette values two different states that she will experi-
H I S TO RY Lionel Robbins (1898-1984)
was a leader in the "ordinal revolution" in
economics. Economics, he wrote, does not
need "to compare the satisfaction which
I get from the spending of 6 pence on
bread with the satisfaction which the Baker
gets by receiving it. That comparison . . . is
never needed in the theory . . . ." (123).
Moreover, "There is no way of comparing the
satisfactions of different people" (124).
160
MICROECONOMICS
- DRAFT
ence. It does not compare her evaluation of a state that she will experience
with someone else’s evaluation of the state they will experience. The first
statement is an example of the cardinal utility that we introduced in Chapter
2 as the basis of expected payoffs (or expected utility) and the analysis of
decision-making in risky situations. The second statement compares Annette’s utility with Brenda’s utility. When utility is represented in this way it is
called inter-personally comparable cardinal utility (or sometimes “cardinal full
comparable utility”). If utility is cardinal in this inter-personally comparable
sense, then we can compare how well off two or more people are, and how
much better off or worse off a policy would make each of them. This provides
a way to evaluate which policies should be implemented by asking whether
H I S TO RY Philosopher-economist John
Stuart Mill (1806-1873) referred to what
we would now call the sum of the total
utilities of a population as as "a good"
that should be promoted: "the general
happiness is desirable... each person’s
happiness is a good to that person, and the
general happiness, therefore, a good to the
aggregate of all persons."4
the gains of those who benefit from a policy exceed the losses of those who
do not.
Why do these two methods of comparing utility matter? Remember that one of
the problems with Pareto efficiency as a criterion for fair policy outcomes was
6
an adequate basis for an Impartial Spectator preferring one outcome over the
5
other. Using the second – stronger – conception of cardinal utility along with
4
the judgement that one outcome is better than another if total utility is greater
provides a rule for evaluating which Pareto-efficient outcome we might prefer
as a society.
Bob's payoffs
many outcomes can be Pareto efficient. So Pareto efficiency does not provide
d
3
c
2
a
1
In the payoffs for the Fishermen’s Dilemma in Figure 1.11 (shown here in
the margin for easy reference) three of the four outcomes of the game are
Pareto efficient. The Pareto criterion provides no way to choose among them.
By contrast the rule — maximize total utility – selects the mutual cooperate
b
0
0
1
2
3
4
5
6
Alfredo's payoffs
Figure 3.17: Pareto comparisons in the Fisherman’s Dilemma Game. The Pareto criterion favors
c over a (which it dominates) but cannot rank
points b, c and d because are all Pareto efficient.
outcome (point c) with total utility of 6.
Cardinal utility and the distribution of wealth
Suppose you are a policy-maker and you have to divide an amount of wealth
between Annette and Brenda. The amount of wealth you have to divide is
equal to 1, so each person can get a fraction of that wealth and, as long
as the fractions sum to 1, then the outcome will be Pareto-efficient. Let the
fraction that Annette gets be a with Brenda’s fraction be (1
a). Annette
and Brenda have identical preferences for wealth given by the cardinal utility
functions of how much wealth they get: uA (a) and uB (1
a). For both of them
the marginal utility of wealth is diminishing with increased wealth.
The horizontal axis in Figure 3.18 shows all possible distributions of wealth
between Annette and Brenda.
• Annette’s share of wealth, a, varies from 0 to 1.
• At a = 0, Annette gets nothing and Brenda gets everything.
E X A M P L E When you say "I’ll do the shopping; it’ll be less trouble for me than for you"
you are representing utility (the trouble of
shopping) as cardinal and making an interpersonal comparison of utility (less trouble
for me than for you). In fact, much of our
everyday ethical reasoning involves interpersonal comparisons of the benefits and costs
that people experience.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
A's marginal utility
uAa
Marginal utility
muB(1 − ah)
Figure 3.18: Distribution of wealth, marginal
utility, and total utility. In the figure a is the
proportion of wealth belonging to Annette (A).
1 a is the proportion of wealth belonging to
Brenda (B). As a person’s wealth increases, the
marginal utility of wealth decreases. A’s total wealth
increases as you move along the bottom line from
left to right, and as a result her marginal utility
decreases. Because B’s wealth increases as the
division moves towards the left, B’s marginal utility
decreases from right to left.
B's marginal utility
− uBa
g
i
h
muA(ah)
0
ai
0
ah
1
A's share of wealth, a
• At a = 1, Brenda gets nothing and Annette gets everything.
Figure 3.18 shows the two marginal utility functions for wealth. Each of them
has decreasing marginal utility in wealth. This means that the increment in
utility associated with each additional unit of wealth they have is less when
they have more wealth. Annette’s marginal utility curve slopes downward
as she gets more wealth (moving from left to right), and Brenda’s marginal
utility of wealth decreases as she gets more wealth (moving from right to
left).
Suppose the status quo is ah , a situation in which Annette is wealthy and
Brenda is poor (Annette’s has share of wealth ah and Brenda’s share of
wealth is 1
ah ). A policy that takes a small amount of wealth from Annette
and transfers it to Brenda reduces Annette’s utility by less than it increases
Brenda’s. We can see this by identifying that Annette’s marginal utility at ah ,
uAa (ah ) is much lower than Brenda’s utility at the same point uBa (ah ). The vertical difference between points g and h shows the magnitude of the difference
in their marginal utilities.
Redistributing wealth from Annette to Brenda therefore increases total utility
(the sum of Annette and Brenda’s utilities).
Applying this reasoning to other points in the diagram, we find that the distribution of wealth that maximizes total utility is ai , where Annette’s marginal
utility of wealth equals Brenda’s marginal utility of wealth. If Annette’s and
Brenda’s utility functions are identical, the total utility maximizing point dis-
161
162
MICROECONOMICS
- DRAFT
tributes wealth equally, ai = 12 .
M-Note 3.13: Maximizing Total Utility
Consider a society of two people, Annette (A) and Brenda (B), in which maximizing total
utility, U = uA + uB , is the goal of the policy-maker who we will assume is Adam Smith’s
Impartial Spectator . The Impartial Spectator selects a point, i, for their choice of policy.
We assume Annette and Brenda assign the same utility numbers to some given level of
wealth. They are identical in this respect. But they differ in their wealth. Annette’s share
of total wealth is a > 12 and Brenda’s is 1 a < 12 < a. Annette’s utility is uA = u(a) and
Brenda’s utility, uB = u(1
a), is also a function of a, but Brenda’s utility decreases as a
increases.
U (a)
uA + uB = u(a) + u(1
=
a)
(3.27)
To find the maximum total utility, we differentiate the total utility with respect to a and set
the derivative equal to zero:
dU
da
) ua (ai )
Ua (a) =
=
uAa + uBa = ua (a)
=
ua (1
=
a) = 0
ai )
The only way that ua (ai ) can be equal to ua (1
ber.
ai
ua (1
1
ai ) is if ai and 1
ai =
ai are the same num-
1
2
This maximization is depicted in Figure 3.18: Annette’s marginal utility, uA
a (a), decreases
as a increases (diminishing marginal utility) and Brenda’s marginal utility uB
a (1
creases as a increases (because her wealth, 1
a) in-
a decreases as a increases). The total
utility maximizing choice occurs at ai , where wealth is equal and as a result the marginal
utilities are equal: uA (ai ) = uB (1
3.14
ai ).
Application: Cardinal utility and subjective well-being
A century ago economists thought that while ordinal comparisons like better or
worse are possible empirical interpersonal comparisons expressed by a number indicating the degree of preferences were impossible to make. But today
researchers are actively engaged in measuring individual happiness and life
satisfaction, using techniques ranging from surveys and natural observation to
E X A M P L E Daniel Kahneman, a psychologist and Nobel Laureate in economic,
has advocated a hedonistic (meaning concerning pleasure and pain) theory of utility.
Kahneman titled one of his papers “Back
to Bentham?" to pay homage to the early
19th century philosopher economist Jeremy
Bentham’s utilitarian theory.5
the methods of experimental neuroscience. They are asking such questions
as: "How important is income for happiness?" "Is being without a job a bigger
source of unhappiness than being without a spouse?" These researchers refer
to happiness or life satisfaction as subjective well-being.
To measure "pleasures and pains" in the lab, volunteers are exposed to an
electrical shock and asked to report on their experience of that on a numerical scale. Others are asked to plunge their hands into extremely cold water
for as long as they can stand it and immediately report their level of unhappiness having done so. Respondents in surveys are asked their "life satisfaction."
This research has sought to understand the activities that make people most
F AC T C H E C K The Satisfaction with Life
survey is based on five questions each
of which is rated on a 7-point scale from
Strongly Disagree (1) to Strongly Agree (7).
Here are the questions: In most ways my life
is close to my ideal; The conditions of my life
are excellent; I am satisfied with my life; So
far I have gotten the important things I want
in life; and If I could live my life over, I would
change almost nothing.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
163
happy. Almost all people surveyed seem to like sex quite a lot, ranking "intimate relations" as having a high subjective well-being value. Ranked after
sex, people like socializing, relaxing, sharing meals with friends, praying, and
exercising. People don’t like housework, childcare, commuting or working.
People also report major changes in subjective well-being from painful events,
like sudden loss of a job, a death in the family, or divorce, or from positive
events like marriage, or the birth of a child.6 But when you ask someone about
their happiness over time the measures are surprisingly consistent: people
are likely to report similar activities or outcomes as providing them with happiness when you ask them at different intervals.
What are the take-home messages about subjective well-being, the choices
about how we spend our time, and how we value effortful work?
First, people like a diverse array of activities and doing things they like provides them with happiness that can be measured in meaningful ways across
different people and at different periods of time in our lives.
F AC T C H E C K Non-laboratory measures of
subjective well-being suggest that people
with higher subjective well-being tend to
be less likely to contract a cold virus and to
recover more quickly when they do contract
the cold. Similar evidence exists for people
who have recovered from wounds and had
baseline and subsequent subjective wellbeing measured: those who are happier
recover more quickly.7
Second, people who report greater subjective well-being are also better off
by physical biological measures. For example,they are less likely to be ill.
Subjective well-being also manifests in hormone levels, brain patterns, and
palm temperature.8
Third, while income matters for happiness (especially for people without much
income) people value social relationships – marriage, a job, friendships –
more than they value income.9 Making the transition from unemployed to
employed boosts a persons subjective well-being by much more than would
be predicted simply by the increase in income. This is because having a job
is a source of respect and dignity, especially as it provides a way for people
to express autonomy over themselves, competence in their expression of
their abilities, and relatedness to other co-workers and people around their
work.
Checkpoint 3.12: Joy or Misery?
Think about the kinds of activities that Kahneman and Krueger discuss above
that provided people with joy (that they ranked highest in terms of providing
them with subjective well-being).
1. Compare them with their opposites: those that result in disutility or even
misery.
2. Come up with a list of activities that you engage in that provide you with joy
which you try to prioritize.
3. Why do you spend the time that you do on these activities? Why do you not
spend more?
4. Do you engage in activities that in the moment are unpleasurable but which
you believe provide you with benefit nonetheless?
E X A M P L E The substantial subjective cost
that people experience when they are out
of work is one reason why employers (who
have the power to terminate a person’s job)
have power over their employees. We shall
return to this when we study the firm and the
labor market.
164
MICROECONOMICS
- DRAFT
5. Do you think such activities appear in the models we’ve developed?
3.15
Preferences, beliefs, and constraints: An assessment
Many scholarly disciplines in addition to economics are devoted to understanding human behavior including psychology, sociology, anthropology and
history, but also more distant endeavors including literature, philosophy, neuroscience, computer science and biology. The preferences, beliefs and constraints approach, while a standard set of tools in economics that is widely
used in other fields, is just one of many approaches. People newly familiar
with the approach often raise the following questions about it.
• Are people really all that selfish? This concern is based on a misunderstanding of the model,which says nothing about whether people are seeking to help others, aggrandize themselves, or a little of both. Our treatment
of altruism, reciprocity and fair-mindedness shows that the model – using
indifference curves and feasible sets, for example – can apply to a variety
of motives.
• Do people consciously optimise, for example, applying the mrs = mrt rule
when they shop? The model is not a description of how people actually
think or their emotional states when they take a break from studying, or
support a particular environmental policy. We model instead what people
would do if they did the best that they could. The fact that the model often yields predictions similar to what we observe empirically (including by
experiments, econometric and other quantitative methods) does not require that the model is an accurate representation of the process by which
people come to take one course of action over another.
In some cases, people consciously optimize, going though mental calculations
similar to the model. For example, a person buying a house or choosing between two job offers will weigh the pros and cons of the alternatives. But in
other cases, the actions may not even appear to us as a decision, for example, what to eat for breakfast, what to wear today, or what our personal values
should be. Without consciously trying to do so, people may arrive at something like the solution to these optimization problems by trial and error, or by
observing others who seem to be successful or happy with their choices, or by
following habits that will remain in place unless changed by some dramatically
adverse consequences of following them.
Other concerns about the model are more serious.
• What about emotions and visceral reactions, aren’t they important? This
question points to a shortcoming of the approach; but it is not that the approach excludes emotions like fear, shame, and attraction. The shortcoming is that the preferences, beliefs, and constraints approach says almost
H I S TO RY In his 1953 work,Essays in
Positive Economics, Nobel Laureate Milton
Friedman (1912-2006) observed that
"predicting the shots made by an expert
billiard player" could be done on the basis
of "the complicated mathematical formulas
that would give the optimum directions" of
the shots. But this prediction would not be
"based on the belief that billiard players,
even expert ones, can or do" actually make
these calculations.10
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
165
nothing at all about motives, that is, it says nothing about the reasons why
people rank some outcome as superior to another. Knowing more about
motives like this would help us understand economic and other behavior.
• Preferences and beliefs are not "just there" as facts of nature, they are
products of environments we live in. We have already seen an example
of this as social norms concerning the kinds of work that are appropriate
for women to do changed during the 20th century under the influence of
new technologies for cooking cleaning and washing and the experience
of women doing factory work producing armaments during the second
world war. Our preferences and beliefs are to some degree "socially constructed". Our discussion of differing cultures around the world using
experiments designed using the preferences, beliefs and constraints framework shows that the approach can help to clarify how cultures differ and
how society shapes preferences and beliefs.
• Commitments and consequences. The framework is based on the idea that
our behavior is based on our beliefs about the consequences our actions
will bring about in the future. Don’t we sometimes act to fulfill promises or
other commitments made in the past, or just to "do the right thing" without
regard to future consequences? Yes we do, and a shortcoming of the
approach is that it does not address that kind of behavior.
• Predicting behavior and evaluating outcomes. Economists use the same
concept "utility" in models designed to predict the actions that people will
take and to provide the basis for evaluating economic outcomes and public
policies to improve them. The idea is that whatever it is that motivates
people to make the choices they do should also be the objective of public
policy and form the basis for our preferring one societal outcome over
another. But treating actual behavior as if it were the pursuit of a concept of
well-being that should be the basis of our judgement of societal outcomes
is a mistake. The reasons for our actions (that is, our preferences) include
addictions, weakness of will, shortsightedness, and other well documented
socially dysfunctional aspects of human behavior that in retrospect are
often deeply regretted by those acting on them.
A sensible conclusion from reviewing these concerns about the preferences
beliefs and constraints approach might be that the approach is better for
answering some questions than others, and learning to distinguish which is
which is an important learning objective. As we said at the beginning of the
chapter: the map is not the territory. Good maps don’t have all the information
about the territory they depict and good economic models require us to leave
some things out.
Checkpoint 3.13: Positive and normative uses of "utility"
H I S TO RY : P O S I T I V E A N D N O R M AT I V E
E C O N O M I C S The distinction between
the economics of "what is" called positive
economics and "what ought to be" called
normative economics was made by John
Maynard Keynes in his 1893 Scope and
Method of Political Economy and by Milton
Friedman in his 1953 The Methodology
of Positive Economics. The distinction is
controversial in part due to differences about
the appropriate role for "what ought to be"
statements in economics.11
166
MICROECONOMICS
- DRAFT
Consider the statement by J.S.Mill (in the above margin note): "each person’s
happiness is a good to that person, and the general happiness, therefore, a
good to the aggregate of all persons." Explain how Mill is here using "happiness"
both as a way of predicting behavior (sometimes called "positive economics")
and as a way of evaluating outcomes from a societal standpoint ("normative
economics.")
3.16
Conclusion
In this chapter we have studied the constrained optimization problems shown
in Table 3.1. Though the problems concerned are quite different, the models
and analytical tools we used to analyze them are very similar. In each case
the analysis of the decision involves two kinds of trade-offs:
• The first trade-off that appeared in each of these situations is the actor’s
relative valuation of the things she cares about, measured by the negative
of the slope of an indifference curve, that is, the marginal rate of substitution.
• The second trade-off is that at any point on the feasible frontier, the opportunity cost of having more of one good that the actor values is that she
must have less of another good that she values. This opportunity cost
trade-off is measured by the negative of the slope of the feasible frontier,
that is, the marginal rate of transformation.
The result - the action taken doing the best she can under the constraints she
faces - is determined in the same way in all the cases: by finding the point on
the feasible frontier that is on the highest indifference curve. This will often be
the bundle where the mrs = mrt rule holds. The table demonstrates that many
seemingly different kinds of action can be studied with a common model, one
that we will use often.
In this chapter, we have focused on single actors and for the most part excluded from the model something important: other people. With the exception
of the farmers of Palanpur, we have modeled the person facing a given situation defined by a feasible frontier and preferences represented by indifference
curves. (We already explained that the second person in the Dictator Game is
not really a player at all).
We now turn to a world populated by people interacting strategically, and
we ask how economic institutions affect the outcomes of these interactions,
and how these outcomes and institutions might be judged by the standards
of Pareto-efficiency and fairness. We shall continue to employ the tools of
constrained utility maximization: we will continue to understand people’s tradeoffs through their marginal rates of substitution and of transformation; and we
shall need these to understand how one person might engage in exchange of
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
Actor
Utility depends on
Action
Constraints
Keiko
Learning, Living time
Time allocation
Learning-Living feasible frontier
Keiko
Learning, Study effort
Study effort
Study-Learning production function
Annette/Chen
Payoffs to two players
Payoffs to each
The total endowment
Annette/Brenda
Conventional goods,
Conventional goods
Consuming conventional goods
Environmental quality
167
degrades the environment
goods and services with someone else.
Making connections
Strategic and non-strategic social interactions: In the previous chapters we
considered strategic social interactions – like the fishers and the farmers
from Palanpur. Here we look at simpler aspects of behavior when a person
is attempting to do the best they can in situations that are not strategic
because the choice of how hard to study, or how much fish to buy is not
greatly affected by others choices.
Self-regarding and social preferences: In Chapter 2 we provided evidence
that people can be self-regarding, altruistic, reciprocal, and fair-minded.
These diverse behaviors can be modeled using the preferences, beliefs,
and constraints approach using indifference curves and feasible frontiers,
as we showed for the case of an altruist.
Opportunity costs and trade-offs: Regardless of whether a person’s preferences are self-interested or social, people face trade-offs among the ends
they wish to pursue and they face opportunity costs when trying to choose
a course of action.
Public policy: Economics engaged: The idea of constrained utility maximization illustrated the trade-off between consuming more goods on the one
hand or either consuming less and using some of the economy’s resources
to abate environmental damages, obtaining greater environmental quality.
We also modeled the choices an altruistic person might make in sharing
something of value thereby providing a model capable of analysing the
kinds of result observed in the experiments reviewed in the previous chapter.
Evaluating outcomes: Treating utility as cardinal and inter-personally comparable rather than ordinal allows us to compare the benefits and burdens
that a policy will impose on different people. This provides a basis (one of
a number of alternatives) for saying that one policy or outcome might be
preferred to another, as illustrated by the case of the distribution of wealth.
Table 3.1: The constrained optimization problems
used in this chapter
168
- DRAFT
MICROECONOMICS
Important ideas
preference
constraints
beliefs
Homo economicus
altruist
reciprocator
ordinal utility
cardinal utility
utility function and value function
total utility
marginal utility
Cobb-Douglas utility
mrs
law of diminishing marginal utility
mrt rule
slope
indifference curve
marginal rate of substitution
diminishing marginal rate of substitution
tradeoff
willingness to pay
iso-value curve
feasible/attainable
feasible frontier
production function
marginal product
marginal rate of transformation
increasing marginal rate of transformation
opportunity cost
increasing opportunity costs
utility-maximizing
point of tangency
price line
offer curve
Mathematical notation
Notation
Definition
u()
x
y
h
a
ȳ
c()
a
p
u
z
p
utility function
a good (or a "bad")
a good (or a "bad")
hours of work
Cobb-Douglas exponent of good x
vertical intercept of the feasible frontier
opportunity cost
A’s share of wealth
price of a good
constant utility along an indifference curve
endowment in the Dictator game
payoff in the Dictator game
Note on super- and subscripts: A, B, C, D: different people; CD: CobbDouglas utility function; Subscript b indicates where someone does the best
they can; RD: feasible frontier with Research and Development.
Discussion questions
See supplementary materials.
Problems
See supplementary materials.
D O I N G T H E B E S T YO U C A N : C O N S T R A I N E D O P T I M I Z AT I O N
Selected Answers/Hints for Questions
See supplementary materials.
169
4
Property, Power & Exchange: Mutual Gains & Conflicts
DOING ECONOMICS
[T]he efforts of men are utilized in two different ways: they are directed to the
production or transformation of economic goods, or else to the appropriation of
goods produced by others.
Vilfredo Pareto, Manual of Political Economy (1905) (Pareto (1971):341)
Ibn Battuta, the fourteenth century Moroccan scholar, reported that along
the Volga River in what is now Russia, long distance trade took the following
form: “Each traveler ... leaves the goods he has brought ... and they retire
to their camping ground. Next day they go back to ... their goods and find
opposite them skins of sable, miniver, and ermine. If the merchant is satisfied
with the exchange he takes them, but if not he leaves them. The inhabitants
then add more skins, but sometimes they take away their goods and leave the
merchant’s. This is their method of commerce. Those who go there do not
know whom they are trading with or whether they be jinn [spirits] or men, for
they never see anyone."
The Greek historian Herodotus describes similar exchanges between Carthaginian
and Libyan groups in the 5th century B.C. After having left their goods,
Herodotus reports, the Carthaginians withdraw and the Libyans “put some
This chapter will enable you to do the
following:
• Explain why, when people exchange
goods, there are both mutual gains and
also conflict over the distribution of these
gains.
• Understand how an allocation of goods
can be evaluated on grounds of Paretoefficiency and fairness.
• Show how self-regarding as well as
social preferences of parties to an
exchange can affect the outcome, and
how other-regarding social preferences
may reduce the scope of conflicts
over the distribution of the gains from
exchange.
• Understand how both private property
rights and the exercise of power by one
of the parties to an exchange will affect
the outcome of exchange.
• Use the ideas you have learned to explain how an employer and an employee
might bargain over working hours and
wages.
gold on the ground for the goods, and then pull back away from the goods. At
that point the Carthaginians ... have a look, and if they think there is enough
gold to pay for the cargo they take it and leave.”
Herodotus describes how the process continues until an acceptable is price
hit upon, remarking with surprise that “neither side cheats the other ... [the
Carthaginians] do not touch the gold until it is equal in value to the cargo, and
[the Libyans] do not touch the goods until the Carthaginians have taken the
gold.”
Alvise da Ca da Mosto, a fifteen century Venetian working for the Portuguese
crown, reported a similar practice in the African kingdom of Mali, regarding it
as “an ancient custom which seems strange and hard to believe.”
Figure 4.1: A painting of Ibn Battuta (on
the right) (1304-1369) Source: Wikimedia
Commons.
172
MICROECONOMICS
- DRAFT
But is the so-called ‘silent trade’ really so odd? Transfers of goods among
strangers can be dangerous. What one expected to be an exchange at mutually agreeable prices may end up as theft or an “offer you cannot refuse.” But
trade among strangers can also be highly profitable. The potential gains from
trade are often greater the more distant geographically or socially the parties
are to the exchange: the salt brought by Tuaregs from the Atlas Mountains in
North Africa across the Sahara by camel to the Kingdom of Ghana was not
available in West Africa. The gold and tropical nuts Tuaregs gained in silent
trade with Ghanaians was not available north of the Sahara.
The silent trade – with its unusual etiquette in which parties interacted only at
a distance – allowed both Tuaregs and Ghanaians to get some of what they
lacked and wanted in return for giving up some of what they had in abundance
and could readily part with.
They were exploiting the mutual gains that differences in geography, tastes,
technologies, and skills allow. And the rules of the game for governing their
exchange process – the institutions that we call "the silent trade" were a
way of doing this and dividing the mutual gains without resorting to violent
conflicts.
Other than these mutually advantageous exchanges, there are many other
ways that goods change hands: from the use of violent coercion by private
parties (i.e. theft), or by the use of one nation’s military force to acquire the
resources of another people. People have also been violently coerced into
work through enslavement by private actors and states alike.
A key characteristic of these coerced transfers is that they are not motivated
by mutual gain, but instead by the gain of one party facilitated by superior
force and institutional power. These transfers of resources and lives have
shaped the course of history and have had important economic consequences
and enduring legacies.
But here we set aside the use of physical coercion and ask how societies
organize the process of exchange motivated not by fear of physical harm but
instead by the prospect of mutual gain. We also provide terms that allow us to
evaluate some of these outcomes as better or worse than others.
4.1 Mutual gains from trade: Conflict and coordination
In a modern economy we engage in indirect monetary exchange: selling some
of our goods or some of our working time for money and using the money to
purchase the goods we need rather than bartering directly as did the Libyans
and Carthaginians. The principles of barter exchange, where goods are
directly transferred among two parties without the use of money, however illustrate the fundamental considerations behind all types of exchange, including
Figure 4.2: A statue of Herodotus. Considered by
many to be the first historian, Herodotus lived in the
fifth century BCE. Source: Wikimedia Commons.
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
indirect monetary exchange.
We will simplify by thinking about just two people who exchange goods directly
with each other, thereby modifying the goods that they hold. To do this we will
introduce two terms describing the bundles that each has before and after
173
R E M I N D E R As in Chapter 3 a bundle is just
a list of the quantity of the goods (or other
thing of value) that a person has. We refer to
the bundles held by all of those involved in
an exchange as an allocation.
exchange:
• The endowment bundle or endowment, the quantities of goods a person
has before exchanging goods.
• The post-exchange bundle the bundle a person has after exchanging
goods with another person.
The bundles held by each of the people (either before or after exchange) is
called an allocation.
Voluntary exchange: mutual gains and conflict over their distribution
An exchange is voluntary if all parties to the exchange have the option to not
engage in it but instead choose engage in the exchange. So each party must
expect to be better off, or at least is no worse off, as a result of the exchange,
which implies that each prefers (at least weakly) their post-exchange bundle to
their endowment bundle.
Recalling the meaning of a Pareto-comparison, we can see that if an exchange is voluntary for both parties, the post-exchange allocation must be a
Pareto-improvement over the endowment, otherwise one or both of the parties
would have refused to participate in the exchange. The stipulation that the
A LLOCATION The bundles held by each of
the people (either before or after exchange)
is called an allocation.
VOLUNTARY E XCHANGE An exchange is
voluntary if all parties to the exchange have
the option to not engage in it but instead
choose engage in the exchange. So each
party must expect to be better off, or at least
be no worse off, as a result of the exchange,
which implies that each prefers (at least
weakly) their post-exchange bundle to their
endowment bundle
R E M I N D E R : An economic rent is the difference between a player’s fallback payoff and
the payoff (profit or utility) they obtain from
participating in an interaction. The gains
from exchange from an interaction is the sum
of the economic rents of all participants.
in order for an exchange to be called voluntary, the post exchange allocation
must be a Pareto-improvement over the endowment bundle is termed the
voluntary transfer requirement.
To make the idea of voluntary exchange concrete we often let the fallback
position of the players be a bundle of goods that is their private property which
they are free to dispose of in exchange or by gift to others, or to retain for
themselves, excluding others.
Let’s review some of the terminology from earlier chapters and explain how
they are used to study the process of exchange.
• A person’s fallback position is what they experience in the absence of the
exchange and the utility number they assign to that bundle (that is, the
utility of their endowment bundle which is considered to be her next best
opportunity.)
• The improvement in utility enjoyed by a party to an exchange is their rent
resulting from the exchange, namely, the difference in utility associated with
their post exchange bundle compared to their fallback position.
P RIVATE PROPERTY Private property is the
right to exclude others from the goods one
owns, and to dispose of them by gift or sale
to others who then become their owners.
VOLUNTARY TRANSFER REQUIREMENT The
stipulation that in order for an exchange
to be called voluntary, the post-exchange
allocation must be a Pareto-improvement
over the endowment bundle is termed the
voluntary transfer requirement.
174
MICROECONOMICS
- DRAFT
• The total rents received by parties to an exchange, also termed the gains
from trade are the utilities of the exchanging parties at the outcome of the
exchange minus the utilities at their fallback positions. .
?? The fact that an exchange is voluntary does not mean that is is fair. Some
exchanges take place under conditions such that one party gains virtually
all of the available rents. How the economic rents are divided between par-
P RIVATE PROPERTY Private property is the
right to exclude others from the goods one
owns, and to dispose of them by gift or sale
to others who then become their owners.
ticipants is the distributional outcome of the exchange. The rents may be
captured by one party, leaving the other with a different set of goods than her
endowment but no better off.
Or the rents may be split among the parties in a way that appears fair, or at
least acceptable to both, as in the silent trade between the Carthaginians
and the Libyans described by Herodotus. The division of the gains from
exchange in the form of economic rents is parallel to the division of the pie in
the Ultimatum Game of Chapter 2.
D ISTRIBUTIONAL O UTCOME How the gains
from exchange – the economic rents –
are distributed between the people in an
exchange; the share of the gains from
exchange each player gets as a rent.
Exchange therefore has two aspects: mutual benefit and conflict of interest:
• Mutual benefit is possible because participants move from their endowment bundle to the post-exchange allocation where they share the gains
from exchange and obtain an economic rent.
• A conflict of interest is present because the gains from exchange can be
divided in many ways among the parties who find themselves in conflict
over who gets the larger share.
Institutions and social norms govern the process of exchange that leads both
to the re-allocation of goods, and to the distribution of the gains from trade.
We will see that institutions and social norms have effects on:
• Pareto efficiency, facilitating or obstructing the realization every opportunity
for mutual gain among the parties to an exchange, and
• The fairness of the distributional outcome, favoring one party or the other in
the conflict of interest in the distribution of the economic rents.
A major institutional challenge today is to find rules of the game that will have
as the Nash equilibrium allocations that are both Pareto efficient and fair. We
will return to the interplay of these two objective frequently in the pages that
follow.
4.2 Feasible allocations: The Edgeworth box
Lets consider a concrete setting in which two people might consider alternative possible distributions of two goods amongst themselves. Let’s say that
H I S TO RY The Edgeworth box is named
after the British economist Francis Ysidro
Edgeworth (1845-1926) who is credited with
having invented this clever way to represent
exchange and bargaining.
P R O P E RT Y, P OW E R ,
8
7
B's coffee (kilograms), xB
6
5
4
3
2
Biko
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
z
0
1
2
Ayanda
3
4
5
6
7
A's coffee (kilograms), xA
8
9
10
10
9
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
8
7
B's coffee (kilograms), xB
6
5
4
3
2
1
Biko
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Better
for Ayanda
Better
for Biko
0
1
2
Ayanda
(a) An allocation, z
Ayanda and Biko have to divide a total of 10 kilograms of coffee and 15 gigabytes of data between them. At the start, nobody owns the goods, the two
quantities are simply amounts available to the two of them. Ayanda and Biko
might now ask each other: what allocation of the coffee and data between the
two of us would be the best?
We use the notation x̄ = 10 and ȳ = 15 to stand for the total amount of coffee
(x) and data (y) available. We define xA , and yA as the quantity of goods
x (coffee) and y (data) in Ayanda’s bundle, and similarly xB and yB are the
quantities in Biko’s. The amount of the two goods in their respective bundles
can be anywhere from zero to the entire amount available, namely, x̄ and ȳ.
Then, an allocation is a particular assignment of coffee and data to the two
people that we can write as (xA , xB ; yA , yB ). An allocation is feasible if the
amounts of coffee and data it gives to Ayanda and Biko is no greater than the
amount available:
xA + xB  x̄
yA + yB  ȳ
Figure 4.3 (a) represents the total supply of the goods, with width and height
equal to the total amount of coffee (x) and data (y) available. The box’s width
is the total amount of x, x̄ (kilograms (kg) of coffee) and its height is the total
amount of y, ȳ (gigabytes (gb) of data). We measure A’s allocation, (xA , yA )
from the lower left-hand corner of the box, and B’s allocation, (xB , yB ) from the
upper right-hand corner.
Any point in the box (or on its edges) is a bundle representing a feasible
allocation of the two goods between the two parties, with the property that it
fully exhausts the total supply of the two goods. You can see this because
the width of the box is the total amount of x and the height of the box is the
3
4
5
6
7
A's coffee (kilograms), xA
8
175
9
B's data (gigabytes), yB
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
1
A's data (gigabytes), yA
9
B's data (gigabytes), yB
A's data (gigabytes), yA
10
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
10
(b) Indifference curves
Figure 4.3: Feasible allocations that exhaust
the supply of both goods. Figure 4.3a shows an
example of a feasible allocation at point z. Figure
4.3b shows the direction in which each person
prefers to move to increase their utility. When
indifference curves are plotted in this rectangle the
graph is called an Edgeworth box.
176
MICROECONOMICS
- DRAFT
total amount of y. Allocation z, for example, gives Ayanda 9 kilograms of
coffee and 1 gigabyte of data and Biko 1 kilogram of coffee and 14 gigabytes
of data (exhausting the 10 units of x and the 15 units of y). There are also
many feasible allocations of the two goods that are not shown in the box. For
example, if Ayanda and Biko each got 1 kilogram of coffee and one gigabyte of
data, that would be feasible given the total amounts, but it could not be shown
in the Edgeworth box because the Edgeworth box only shows allocations
where the two people divide up all of the goods so that they sum to x̄ and
ȳ.
As we move to the northeast in the box, Ayanda gets more of both goods,
and as we move to the southwest in the box, Biko gets more of both goods.
Because both are self-regarding we show this on the figure with the arrows
labeled: "Better for Ayanda" and "Better for Biko" respectively .
How can we evaluate whether some allocations are better than others? To do
this we can represent the preferences of the two parties by plotting their indifference curves in the box. This allows us to say for both Ayanda and Biko that
for any two allocations (points in the box) that the first is preferred to the second, the second is preferred to the first, or the person is indifferent between
the two. To do this we need to know the utility functions of the two.
Both Ayanda and Biko enjoy consuming both coffee and data. Their utility
functions are:
Ayanda’s utility function
uA (xA , yA )
Biko’s utility function
uB (xB , yB )
We assume that the indifference curves for both parties exhibit decreasing
marginal utility for both goods. To provide a concrete example, we will assume
that both Ayanda’s and Biko’s utility functions are Cobb-Douglas, but in some
cases that follow, with different preferences for coffee and data:
A
aA)
B
aB)
Ayanda’s utility function
uA (xA , yA ) = (xA )a (yA )(1
Biko’s utility function
uB (xB , yB ) = (xB )a (yB )(1
In numerical examples we will often contrast two cases:
• Identical: The two people have identical preferences for the two goods,
such as a A = 12 , a B = 12 .
• Different: The two people have different preferences, for example, such
that A’s a A = 23 , whereas for B a B = 13 . So Ayanda has a stronger preference for coffee than Biko does.
We can visualize the allocation of coffee and data between Ayanda and Biko
using an Edgeworth box. An Edgeworth box allows us to see both people’s indifference curves in the same space to identify mutually beneficial
trades.
R E M I N D E R Recall that in Chapter 1, we
used z to indicate the fallback position of
people playing games in the general form of
the Fishermen’s Dilemma game with ranked
outcomes. At z, the people experience their
B
utilities, uA
z and uz , as their utility at the
fallback position, that is, their endowments if
they do not trade.
15
15
14
14
13
13
12
12
uA3
10
z
11
10
A's data (gigabytes), yA
uA2
B's data (gigabytes), yB
uA1
11
●
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
10
9
8
7
6
5
4
3
9
8
7
6
5
uB3
4
3
2
uB2
2
z
1
uB1
1
0
0
0
Ayanda
1
2
3
4
5
6
7
A's coffee (kilograms), xA
8
9
10
(a) Ayanda’s Indifference Curves
0
1
2
3
4
5
6
7
8
9
10
B's coffee (kilograms), xB
Biko
(b) Biko’s Indifference Curves
Ayanda and Biko’s indifference curves are shown separately in Figure 4.4
panels a and b. In panel c. we plot the same indifference curves together in
the Edgeworth box. Ayanda evaluates the allocations from the point of view of
the lower left-hand corner, and her indifference curves represent higher utility
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
uA1
9
8
B's coffee (kilograms), xB
7
6
5
4
3
2
177
Biko
1
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
uA3
uA2
uB3
uB2
uB1
z
0
Ayanda
1
2
3
4
5
6
7
A's coffee (kilograms), xA
8
9
B's data (gigabytes), yB
A's data (gigabytes), yA
P R O P E RT Y, P OW E R ,
10
(c) Indifference curves in an Edgeworth Box
Figure 4.4: Indifference curves and an Edgeworth box. In panels a. and b. we show three of
Ayanda’s and Biko’s indifference curves respectively. In panel c., Biko’s indifference curves have
been flipped 180°so that the origin in the lower left
of panel b. has become the origin of the Edgeworth
box at the upper right.
as we move to the northeast in the box.
Ayanda’s indifference map looks exactly the same in the Edgeworth box as
it does in the separate plot, because in both cases the origin from which we
measure her allocation is in the lower left-hand corner. Biko evaluates the
allocations in the box, however, from the point of view of the upper right-hand
corner, and his indifference curves represent higher utility as we move to the
southwest in the box. It may help you understand how we superimposed
Biko’s preferences on Ayanda’s if you think about what we called their "point
of view." In panel’s a. and c., imagine Ayanda standing at the lower left origin
and looking up her indifference map, as if the curves were contours of a
mountain, the curves farther away being at higher altitudes. Now do the same
with Biko, but for him when he looks to the north east in panel b., he is looking
R E M I N D E R In Chapter 3 we defined the
Cobb-Douglas (CD) family of utility functions
as:
u(x, y) = xa y(1
a)
(with 0  a  1). The Cobb-Douglas
utility function results in a marginal rate of
y
substitution, mrs(x, y) = (1 aa ) x .
up his "utility mountain." But in panel c. he is standing at the upper left origin
and the way up his utility map is to the south west.
In the figures, at allocation z Ayanda and Biko have allocations (xzA , yA
z) =
(9, 1) and (xzB , yBz ) = (1, 14). The indifference curves that go through allocaA
B
B
tion z provide Ayanda and Biko with utilities uA
z = u2 and uz = u2 .
A
In panels a. and c., uA
2 = uz is Ayanda’s indifference curve through z. In panB
els b. and c., uB
2 = uz is Biko’s indifference curve through z. The indifference
maps for both Ayanda and Biko have indifference curves through every point
in the box, but (following "the map is not the territory" principle) we show only
three in the figure.
M-Note 4.1: Evaluating utilities at an allocation
Given the assumption of that Ayanda’s utility function is a Cobb-Douglas with a A = 23 and
Biko’s utility function is a Cobb-Douglas with a B = 13 , we can calculate their utilities at the
allocation z. Remember that the exponents in the Cobb-Douglas utility function represent
M - C H E C K Biko’s indifference map would
look exactly the same as in Figure 4.4 b.
if we rotated the Edgeworth box 180°to
measure Biko’s allocation from the lower
left-hand corner.
178
MICROECONOMICS
- DRAFT
the person’s intensity of preference for the good. In this example, Ayanda likes coffee
more than Biko does.
2
1
Ayanda has a Cobb-Douglas utility function uA (xA , yA ) = (xA ) 3 (yA ) 3 :
• She has 9 kilograms (kgs) of coffee and 1 gigabyte (gb) of data.
• So her allocation at point z is (xzA , yA
z ) = (9, 1)
• At her allocation z her utility is uA (xzA , yA
z ).
2
1
• So for 9 kgs of coffee and 1 gb of data: uA (9, 1) = (9) 3 (1) 3 = 4.326749.
Checkpoint 4.1: Biko’s utility at allocation z
Using the method shown in M-Note 4.1, what is Biko’ utility at the allocation
given by point z in the Edgeworth box.
4.3 The Pareto-efficient set of feasible allocations
Which allocations in the Edgeworth box are Pareto-efficient?
It’s easy to see that simply throwing away some of x or y cannot be efficient
because allocating those portions to Ayanda and or Biko instead would have
made at least one of them better off without making the other worse off. So
Pareto-efficiency also requires that Equations 4.1 are satisfied as equalities,
not as inequalities. By construction, any of the great many allocations in
the Edgeworth box allocates all of the coffee and data to one or the other
participant, and meets this criterion.
To narrow things down, Ayanda and Biko could agree that the final allocation
chosen must be Pareto-efficient. In Figure 4.5 we show Ayanda and Biko’s indifference curves through an arbitrary allocation z and three more indifference
curves for each person: two indifference curves higher and one indifference
curve lower than for allocation z.
The endowment allocation is not Pareto-efficient
Think about z as a hypothetical allocation, for example, if Biko said: "Ayanda,
how about you have 9 kg of coffee and I get the 1 kg remaining, while I get 14
gb of the data, and you get the 1 gb remaining." We can see, however, that z
in Figure 4.5 is not Pareto efficient. The reason is that at the allocations given
by point z, Ayanda’s and Biko’s indifference curves:
• intersect, which means
• they have different slopes,
• indicating different marginal rates of substitution
R E M I N D E R In games like the Ultimatum
Game in Chapter 2 any allocation of the pie
in which the entire endowment is allocated
to one of the players or the other – in other
words " no money left on the table" is
Pareto-efficient. But the allocations resulting
from the Ultimatum Game are frequently
inefficient because when the Responder
rejects the Proposers offer both players get
zero, and all of the money is left on the table.
M - C H E C K Even if for some reason we were
not to allow the allocation to involve fractional
quantities of the goods and require that
allocations be integers, there 176 possible
allocations to be exact (that’s 11 ⇥ 16, in
case you are wondering, because we would
then have to include zeroes as possible
allocations for the goods).
R E M I N D E R The marginal rate of substitution
is the negative of the slope of the indifference curve. It is also equal to the ratio of
the marginal utilities of the two goods, x and
y, i.e. mrsA (x, y) =
uA
x
. The marginal rate
uA
y
of substitution is also the willingness to pay
for x in terms of y. The people’s marginal
rates of substitution have the dimensions
data/coffee (data for coffee).
R E M I N D E R For an outcome to be Paretosuperior to another, at least one participant
must be made better off – get higher utility –
and no participant can be made worse off –
get lower utility.
P R O P E RT Y, P OW E R ,
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
9
8
uA1
B's coffee (kilograms), xB
7
6
5
4
3
2
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Pareto−
efficient
curve
uA3
uAz
uB1
uBz
d
uB3
tA
uB4
i
tB
0
Ayanda
1
2
3
4
5
h
6
7
A's coffee (kilograms), xA
z
8
9
Figure 4.5: Pareto-efficient allocations To
2
10
goods, which means their willingness to pay to acquire more of one or the
other good differ;
• and this means that there is a feasible Pareto-improving exchange that has
not been realized,
• so these allocations are not Pareto efficient.
Figure 4.5 provides us with a numerical demonstration of the above logic.
We have seen that the difference between the two people’s marginal rates
of substitution at the point z indicates that they can make a Pareto-improving
trade – Ayanda giving up some of her coffee in return for some of Biko’s data
– at their endowment allocation z.
In M-Note 4.2 we show that Ayanda’s mrs is 29 and Biko’s is 7 in Figure 4.5.
This means that Ayanda is willing to pay at most a kilogram of coffee for 29 of
a gigabyte of data, while Biko is willing to trade at most 7 gb of data for one kg
of coffee.
A mutually beneficial trade could therefore take place at any price of coffee
between 29 of a gb of data and 7 gb of data. The low price would benefit Biko,
with Ayanda not improving her utility at all. Correspondingly, if they traded at
the high price Ayanda would make all the gains.
We return to how the price might be determined later. For now, we can eliminate point z in Figure 4.5 as a candidate for being a Pareto-efficient alloca-
1
make this figure we let uA = (xA ) 3 (yA ) 3 and
• meaning that Ayanda and Biko have different offer-prices for the two
tion.
179
Biko
1
B's data (gigabytes), yB
A's data (gigabytes), yA
10
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
uB =
1
( xB ) 3
2
( yB ) 3
. So Ayanda prefers has a
stronger preference for coffee, and Biko for data
(they have asymmetrical preferences). Allocation
h is Pareto-superior to allocation z, but it is not
Pareto-efficient because an alternative point, e.g.
allocation tB , is Pareto-superior to point h (Biko
is better off without Ayanda being worse off). All
points along the Pareto-efficient curve between i
and tB would be both Pareto-superior to h and z
and Pareto-efficient.
180
MICROECONOMICS
- DRAFT
M-Note 4.2: mrs in the Edgeworth box
At allocation z (9, 1), (1, 14) in Figure 4.5, we can calculate each person’s marginal rate
of substitution and compare them. We computed what a person’s mrs(x, y) is when she
has Cobb-Douglas utility in M-Note 3.4 in Chapter 3. We obtain Biko’s from the same
reasoning. We shall assume for this example that the two have asymmetrical preferences
as in Figure ??.
Let’s start with Ayanda, assuming a A = 23 :
uA
yA
• mrsA (x, y) = uxA = 2 xA
y
1
2
A A
• Substitute in A’s allocation at z: mrsA
z (xz , yz ) = 2 9 = 9
Ayanda is willing to pay 29 of a gigabyte to get a kilogram of coffee, or to sell a kilogram of
coffee for 29 of a gigabyte of data.
Now for Biko, assuming a B = 13 :
uB
yB
• mrsB (xB , yB ) = uxB = 12 xB
y
1 14
B B
• Substitute in B’s allocation at z: mrsB
e (xe , ye ) = 2 1 = 7
Biko is willing to pay 7 gigabytes of data for a kilogram of coffee, or to sell kilogram of
coffee for 7 gigabytes of data.
We can see that mrsA < mrsB because 29 < 7. This shows up in Figure 4.5 where the
slope of Ayanda’s indifference curve is steeper than the slope of Biko’s indifference curve
at allocation z.
Which allocations are Pareto efficient? The mrsA = mrsB rule
The same reasoning allows us to eliminate most of the other points too. Remember the demonstration that showed point z to be Pareto-inefficient started
with "at the allocations given by these points Ayanda’s and Biko’s indifference
curves intersect." We explain this further in M-Note 4.2. So any allocation at
which the indifference curves intersect, like point h in Figure 4.5 cannot be
Pareto efficient.
To find the Pareto-efficient allocations, we need to determine which allocations
remain after we have eliminated all of those at which the indifference curves
cross. To do this we can run the above reasoning in reverse.
If the two indifference curves (one of Ayanda’s, one of Biko’s) share a common
point (that is, that represent the utilities at a particular allocation) but do not
intersect, then the two indifference curves must be tangent. This tells us
(reversing the logic above about indifference curves that intersect) that if
Ayanda’s and Biko’s indifference curves:
• are tangent, this means that
• they have the same slopes, indicating
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
181
• identical marginal rates of substitution,
• meaning that Ayanda and Biko have the same willingness to pay for the two
goods.
• This is the same as saying that their maximum willingness to pay to acquire
more of the other’s good is not greater than the least price at which the
other would part with their good
• and this means that there is no feasible Pareto-improving exchange into
which both would voluntarily enter
• so the status quo allocation is Pareto efficient.
This gives us the following rule for an allocation between two players, A and B,
being Pareto efficient:
The mrsA = mrsB rule:
mrsA (xA , yA ) = mrsB (xB , yB )
(4.1)
This rule differs from the seemingly similar mrs = mrt rule for a single individual because this new rule applies to strategic interactions among two or more
inter-dependent actors, of the kind that occur in markets for labor, credit, and
R E M I N D E R : T H E mrs = mrt RU L E We
derived a similar rule for single person
interactions in Chapter 3 The mrs = mrt
rule (with a few exceptions) identifies
the constrained optimal allocation for a
single individual as the bundle at which the
marginal rate of substitution (the person’s
willingness to pay for more of the y-good) is
equal to the marginal rate of transformation
(the opportunity cost of getting more of the
y-good).
many goods. The superscripts A and B are there to remind you that two (or
more) players are involved in this rule.
The points tA , tB and i lie on the purple Pareto-efficient curve in Figure 4.5.
We will often abbreviate the Pareto-efficient curve to PEC. The Pareto-efficient
curve consists of all Pareto-efficient allocations, including Ayanda getting all of
both goods, or the reverse.
A
M - C H E C K Like the mrs = mrt rule, mrs =
mrsB does not work in every case. The
Confining allocations to the Pareto-efficient curve limits the choices that
Pareto-efficient point may be a corner
solution (not a tangency) at which one of
the goods is not consumed at all by one of
the players, and a tangency identified by the
rule may be a minimum not a maximum. The
reasons are the same as were explained for
the mrs = mrt rule.
Ayanda and Biko need to make. But the question is still far from answered.
Moving from one Pareto-efficient allocation to another must make one of the
participants better off and the other worse off. The Pareto efficiency criterion
is not going to help them decide which of the points on the Pareto-efficient
curve they would consider to be the best.
So they face a problem and a conflict of interest.
• The problem is that there are still innumerable Pareto-efficient outcomes on
the PEC and they need some way to decide which one to choose.
• The conflict of interest is that Ayanda prefers points on the PEC to the
northeast in the Edgeworth box, while Biko prefers points to the southwest,
so they will not agree on which Pareto-efficient division of the coffee and
data to make.
M-Note 4.3: Computing the Pareto-efficient Curve
We will use mrsA = mrsB rule to work out the equation for the Pareto-efficient curve.
PARETO - EFFICIENT CURVE The Paretoefficient curve is all outcomes that are
Pareto-efficient. At a Pareto-efficient outcome the marginal rates of substitution of the
two parties are equal so the mrsA = mrsB
holds. The Pareto efficient curve is sometimes called the "contract curve", a term we
do not use because there need not be any
contract involved (e.g. when an outcome in
our thought experiment was implemented by
the Impartial Spectator.
182
MICROECONOMICS
- DRAFT
To find the Pareto-efficient curve, we set Ayanda’s marginal rate of substitution equal
to Biko’s marginal rate of substitution. We already know that mrsA (xA , yA )
=
B
A
2 xyA and
mrsB (xB , yB ) = 12 xyB . We also know that x̄ = xA + xB = 10, so xB = x̄
xA and
ȳ = yA + yB = 15, so yB = ȳ yA . Solutions to these equations for xA , yA , xB , yB are
Pareto-efficient allocations.
We set the marginal rates of substitution equal to each other and use these conditions to
find the Pareto-efficient curve:
mrsA (xA , yA )
=
mrsB (xB , yB )
=
4(10
xA
yA
4 A
x
xA )yA
=
1 ȳ yA
2 x̄ xA
15 yA
10 xA
xA (15 yA )
40yA
4xA yA
=
15xA
(40
3xA )yA
=
15xA
yA
=
15xA
40 3xA
2
Pareto-efficient Curve
yA
=
xA yA
The Pareto-efficient allocations lie on an upward sloping curve between the two origins.
Checkpoint 4.2: Conflict and symmetry of preferences on the Paretoefficient curve
a. Using Figure 4.5 do the following:
i. Explain Ayanda’s and Biko’s preference among the Pareto-efficient
points tA , tB , and i.
ii. Show that they rank these points in opposite order.
iii. Explain why for any two points on the Pareto-efficient curve, Ayanda will
prefer one point and Biko another point; they will never agree on which
is preferable.
b. Work out the formula for the Pareto-efficient curve when the two people have
identical Cobb-Douglas utility functions where a A = a B = 12 . (It’s easier
than the asymmetrical case.) The solution is that the Pareto-efficient curve
is given by yA = 32 xA . But it’s important for you to work out how to get the
solution.
4.4 Adam Smith’s Impartial Spectator suggests a fair outcome
Ayanda and Biko are going to have to figure out some way – other than each
simply trying to get more – for picking an allocation. This means stepping
back and looking at the problem without thinking about their own particular
preferences. They would probably experiment with some simple rules. They
could adopt:
• "finders keepers" rule and allocate the goods to whoever had first discovered the discarded coffee and data; but this might not seem fair.
• the fifty-fifty norm of the landlords and sharecroppers in Chapter 2, and
E X A M P L E To see how maximizing total
utility might lead to unacceptable outcomes,
think about two people, one who in order
to minimize her carbon footprint or for other
ethical reasons has cultivated a simple life
style and is not much interested in increasing
her material consumption and the other who
has cultivated a taste for luxuries and will
be miserable without them. Maximizing total
utility would require giving most of the goods
to the second person.
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
each take half the quantity of the two goods; but if they have different
preferences (as is the case in panel b of Figure 4.5) splitting both goods
equally would not even be Pareto-efficient (an equal split is not on the
purple Pareto-efficient curve.)
• the maximize total utility principle; but this places no value on equality, and
might result in selecting an allocation in which one person had most of the
goods (and utility) and the other little of either.
To develop more satisfactory rules, they might consult an Impartial Spectator
a fair and impartial spectator who can assist them (and us) in reasoning about
what a good outcome might be. We use upper case letters for her name to
remind you that she is an entirely made-up character, a thought experiment,
and not a part of the game in which Ayanda and Biko are engaged. The Impartial Spectator is not a person, she is a thought experiment representing our
conscience, allowing us to explore differing values and how they would lead us
(and Ayanda and Biko) to select a particular allocation as the best.
We’re going to follow the Impartial Spectator’s thinking by looking at different
criteria that she could adopt. For example, she could ask:
• Are the procedures that determined the allocation fair?
• Is the outcome itself fair?
The first criterion is referred to as a procedural judgement, and therefore
she judges the outcome based on the procedure used to acquire the goods.
She would ask for example if the original endowment bundles had been acquired fairly, for example through hard work, gifts, or exchanges in which both
Ayanda and Biko had an equal opportunity to acquire the goods. She would
go on to inquire if the process of trading had itself been fair: for example did
either of them have unfair advantages in determining the price at which they
would exchange.
The second criterion is called substantive: it asks about the substance of
the resulting allocation, asking for example if it is fair (no matter how it came
about).
Both criteria are important, but we will focus on the substantive judgements
because it allows us to illustrate how the Impartial Spectator could select
the “best" allocation by solving a constrained optimization problem. For the
Impartial Spectator to make judgments among Pareto-efficient allocations
that give Ayanda and Biko different levels of utility using the constrained
optimization method, she needs to refer to two pieces of information:
• The set of all Pareto efficient combinations of utility levels that Ayanda and
Biko could experience by allocating the goods in different ways;
183
184
MICROECONOMICS
- DRAFT
• How she (the spectator herself) values each of these combinations of the
utility levels of the two.
The utility possibilities frontier
Setting aside Pareto-dominated allocations, the Impartial spectator will concentrate on the boundary of the set of feasible utility pairs of the two. This is
called the utility possibilities frontier (UPF) and it, shows all combinations of
Ayanda and Biko’s utilities associated with allocations on the Pareto-efficient
curve.
In Figure 4.6 In panel a we show an Edgeworth box of the two player’s allocation problem in which they have identical preferences in the way they each
value coffee and data.
In panel b, we show the utility possibilities frontier for this case. For the moment, ignore the downward-sloping blue lines.
The utility possibilities frontier shows Ayanda’s utility (uA ) on the horizontal
axis and Biko’s utility (uB ) on the vertical axis as we move from one extreme of
the Pareto-efficient curve in figure 4.5 to the other.
The UPF is downward-sloping because the participants are in conflict over
who gets what share of the possible distributions of utility as the allocation
changes on the Pareto-efficient curve in the Edgeworth box.
The UPF is
constructed from the Pareto-efficient curve by translating each Pareto-efficient
allocation (xA , yA ; xB , yB ) into a point (uA (xA , yA ), uB (xB , yB )) that represents
the utility levels of the two participants at that allocation. To construct it, take
any point on the Pareto-efficient curve in Figure 4.5, say point tA , then read
U TILITY P OSSIBILITIES F RONTIER (UPF)
The utility possibilities frontier is a curve
plotted with uA on the horizontal axis and
uB on the vertical axis that shows the utility
of the two participants at all Pareto-efficient
outcomes.
from the two indifference curves through tA the two levels of utility of Ayanda
and Biko at that allocation (namely 8.52 and 3.74 respectively), then go to
Figure 4.6 where those two utility levels become the coordinates in the utility
possibility graph of point tA in the Edgeworth box graph.
Points tA , i, and tB correspond to the same lettered points in Figure 4.5 and
portray the utilities of each of the two at these Pareto-efficient allocations. In
similar fashion, points z and h correspond to the same letters in Figure 4.5,
but these allocations, being Pareto-inefficient are of no interest to the Impartial
Spectator.
As in the case of other feasible frontiers the negative of the slope of utility
possibility frontier,
DuB
,
DuA
is the marginal rate of transformation of B’s utility
into A’s utility by progressively giving A more of the goods and B less. This is
also the opportunity cost of A having more utility in terms of the sacrifice in B’s
utility necessary to allow this. A steep utility possibility frontier means that for
A to gain one unit of utility, B must sacrifice a lot.
R E M I N D E R The utility possibilities frontier
is similar to what we did in Chapter 1 to
understand the Pareto efficiency of different
game outcomes. The UPF is another
feasible frontier introduced in Chapter 3,
since it shows the feasible combinations of
utility possible given the available goods and
the preferences of the participants.
P R O P E RT Y, P OW E R ,
B's coffee (kilograms), xB
8
7
6
5
4
3
2
Biko
1
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Pareto−efficient
curve
d
uA3
uA1 uAz
A
t
i
tB
uB3
uB4
uBz uB1
h
0
Ayanda
1
2
3
4
5
6
7
A's coffee (kilograms), xA
z
8
9
10
(a) Identical Preferences Edgeworth box
13
185
Infeasible
combinations
of utility
12
11
10
tB
9
B's Utility, uB
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
9
B's data (gigabytes), yB
A's data (gigabytes), yA
10
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
8
mrs = mrt
Impartial Spectator's
iso−social
welfare curves
7
i
h
6
5
4
3
w6
tA
z
w5
w4
w3
w2
w1
Feasible
combinations
of utility
2
1
Utility possibilities frontier
0
0
1
2
3
4
5
6
7
8
A's Utility, uA
9
10
11
12
13
(b) The utility possibilities frontier (UPF) and the Impartial Spectator’s
iso-social welfare curves (w) Figure 4.6: The utility possibilities frontier
Checkpoint 4.3: The UPF and the PEC
1. Explain why the utility possibilities frontier in Figure 4.6 is downward-sloping.
2. Explain why if the utility functions of the two differ, an even split of the two
goods – half of each to Ayanda and half of each to Biko – could not be the
Impartial Spectator’s choice of the best allocation.
The Impartial Spectator’s social welfare function
(UPF) and the Impartial Spectator’s iso-social
welfare curves (w). The utility functions of the two
players used to create panel a are identical, with
in both cases a = 0.5. Because they both value
the two goods in the same way, they consume
them in the same proportions at all points on the
PEC. The only differences is which player has
more. Each point in Panel b. corresponds to an
allocation in the Edgeworth box shown in panel
a. The downward-sloping curves in Figure b are
the Impartial Spectator’s iso-social welfare curves,
corresponding to six levels in his judgement of
social welfare w1 through w6 . Social welfare is
lower at points closer to the origin. The allocation
given by point i is the social optimum determined
by the mrs = mrt rule.
Which point on the UPF – in other words which allocation of the goods between Ayanda and Biko – the Impartial Spectator ranks as best will depend
on her values. She has to compare how much she values the utility of Ayanda
and Biko respectively and how this varies depending on the level of utility that
each are experiencing.
To do this she has to be able to compare how much the levels of utility for the
two for each of the allocations on the UPF. She knows that Ayanda prefers
allocation tA to tB (and that Biko ranks these two allocations the other way
around). She needs to treat the utility of each like ordinary numbers that
measure the size not just the rank of something, in this case the cardinal
utility of each.
A summary of the Impartial Spectator’s evaluation of different utility distributions (uA , uB ) is provided by her social welfare function, W (uA , uB ). This is
similar to the utility function that expresses a person’s preferences over bundles of goods, (x, y), but remember the Impartial Spectator is not a person,
but a thought experiment. This is why it is called the social welfare function
rather than the Impartial Spectator’s utility function. A social welfare function provides a way of treating the utilities of the citizens as being cardinal
S OCIAL W ELFARE F UNCTION A social
welfare function is a representation of "the
common good" based on a some weighting
of the utilities (uA , uB , and so on) of the
people making up the society. We can
write a social welfare function in the form
W (uA , uB ).
186
MICROECONOMICS
- DRAFT
numbers that are comparable across people (like height or weight).
R E M I N D E R Assigning cardinal utility numbers to bundles means that we can make
statements like:
An example is a social welfare function that expresses total welfare as the
• for Annette, the outcome (x0 , y0 ) is twice
as good as (x, y) but also
product of the utility of the citizens, each utility raised to some exponent.
Example Social Welfare Function:
W (uA , uB ) = (uA )l (uB )1
l
(4.2)
This social welfare function has the same form as a Cobb-Douglas utility
function: the participants’ levels of utility are the "goods" for the Impartial
Spectator. When l = 0.5 = 1
l , then the Impartial Spectator:
• the sum of the utility experienced by
Annette and Brenda is greater with
outcome (x0 , y0 ) than with outcome
(x, y) because uA (x0 , y0 ) + uB (x0 , y0 ) >
uA (x, y) + uB (x, y)
The sum of the utilities of the two – in the
second statement – is an example of a social
welfare function.
• weights the two peoples’ utilities equally ; and
• places diminishing marginal value on increases in the utility of either of
Ayanda or Biko; the more they consume of the goods the greater is their
utility and therefore the less they add to the Impartial Spectator’s social
welfare.
Because the Impartial Spectator values the two people’s utilities equally, and
(in the judgement of the Spectator) the marginal value of increased utility is
diminishing, she will not rank highly any outcome in which one or the other
gains most of both goods.
Just as we can use indifference curves to represent a person’s utility function
over goods, we can use iso-social welfare curves to represent the Impartial
Spectator’s social welfare function over the utility distribution between people.
The level of social welfare is the same along an iso-social welfare curve, just
as utility was the same along an indifference curve.
Given the Impartial Spectator’s social welfare function, the problem of choosing the Pareto-efficient allocation of coffee and data becomes a constrained
maximization problem similar to those we studied in Chapter 3. The feasible
frontier for the Impartial Spectator is the utility possibility frontier, because it
represents the levels of utility that are achievable given the amount of goods
available and the preferences of the participants.
The iso-social welfare curves of the social welfare function are analogous to
indifference curves for a single individual, but apply to the utilities of the two
people not the two goods consumed by the single individual and express the
valuations of the Impartial Spectator, not the preferences of the individual.
Similar to the individual indifference curve, the negative of the slope of the isosocial welfare curve at any point (uA , uB ) is the Impartial Spectator’s marginal
rate of substitution of Ayanda’s utility in terms of Biko’s utility. And we can
use the mrs = mrt rule to find the constrained social welfare-maximizing
allocation. It is the point where the UPF is tangent to an iso-social welfare
curve.
I SO - SOCIAL WELFARE CURVE Iso-social
welfare curves show constant or equal
("iso") levels of welfare, W̄ , for different
combinations of utility between A and B.
The negative of the slope of the iso-social
welfare curve is the Impartial Spectator’s
marginal rate of substitution (mrsSW (uB , uA ))
of Ayanda’s utility for Biko’s utility.
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
For example, suppose the Impartial Spectator’s social welfare function
is:
1
1
W (uA , uB ) = (uA ) 2 (uB ) 2
which puts a identical weight on the utilities of the two parties, then given that
Ayanda and Biko have identical utility functions, the social welfare maximum
shown in the Edgeworth box in Figure 4.5 Panel a is xA = 5, yA = 7.5, xB =
5, yB = 7.5 or a fifty-fifty split of each good. In the utility possibility frontier
graph in Figure 4.6 this is point i. If their preferences differed, then the social
optimum would result in each getting different amounts of x and y.
Different Impartial Spectators might have different social welfare functions
to rank distributions of the utilities of the two parties, leading to the choice of
different social welfare maximizing Pareto-efficient allocations. Societies do
not have an Impartial Spectator to determine how to weight the competing
interests of society’s members in a social welfare function. Instead, in a democratic society we debate the question of distribution and sometimes come to
a consensus (and sometimes to a deadlock). Controversy about the rights
and wrongs of economic policies such as the tax rates paid by wealthy people and the provision of public services to all, are often implicitly about the
weights (such as a, in Equation (4.2)) that policy-makers should place on the
well-being of different people.
Here we see a sharp contrast between the Pareto-efficiency criterion and
the maximization of social welfare. Preferring a particular Pareto-efficient
allocation over an alternative allocating in which both are worse off cannot be
a matter of conflict. Maximization of some particular social welfare function
subject to the constrain of the utility possibility frontier – some gaining and
some losing depending on the social welfare function used – and is certain to
be controversial.
The imaginary Impartial Spectator helps us understand how values dictate
what we think of as better or worse allocations. These outcomes, as we have
seen in previous chapters and we will now see in greater detail, depend on the
rules of the game. So the Impartial Spectator will have something to say about
how we evaluate which are better or worse institutions by which organize the
process of exchange.
4.5 Property rights and participation constraints
The scenario of Ayanda and Biko enjoying their coffee and data in their student residence and deciding how to allocate them helps us understand the
abstract issues of Pareto efficiency and fairness. Very similar issues arise
when instead we consider Ayanda and Biko to be total strangers, interacting in
a market. But in this new setting the allocation will not be determined by some
imaginary Impartial Spectator. Instead, the allocation will be determined by
187
188
MICROECONOMICS
- DRAFT
who initially owns which goods and the rules of the game that regulate how
Ayanda and Biko might benefit by exchanging some of their goods with each
other.
Market institutions: Property rights and participation constraints
Nobody actually owned the data and coffee that the Impartial Spectator allocated in our thought experiment and Ayanda and Biko were not really engaged in a game. This is not how markets work. Key aspects of the rules of
the market game are:
• The rule of law establishes that the institutions – the laws and other informal rules – governing the interaction are observed, and not violated by
arbitrary acts (for example theft of the others goods by one of the traders or
confiscating by a government official).
• Private ownership. At any moment in the game the goods are the private
property of one or the other of the players, so a point in the Edgeworth box
indicates a distribution of property between the two. The distribution at the
start of the game is called each player’s endowment.
• Private property and the rule of law mean that each player as option to
refuse offers so any exchanges that a player will agree to participate in
must be Pareto improvements over the endowment
• Asymmetric bargaining power will affect the nature of the exchanges
that are executed, and who captures the greater share of the gains from
exchange.
Private property does not distinguish between the two parties: each have
identical rights to exclude the other from their bundle of goods. This would be
true even if Biko initially owned all of the goods and Ayanda had none or the
other way around. In this respect private property rights provide a level playing
field because the right to exclude others from the use of your goods does not
depend on how many goods you have, or on your identity.
The exchange process begins with the property people "start with," that is,
their endowment allocation. These endowments exist before the exchange
we are considering happens. But we are cutting into time at a particular mo-
H I S TO RY It has not always been true that
one’s property rights did not depend on your
identity. In many societies, some people –
such as women – did not have the right to
own property, and some people – such as
enslaved people – were treated as property.
ment. These endowments, which are the status quo of our game, are the
result of similar games played in the past, and also other games in which who
owns what goods may have been determined by force and not by voluntary
exchange. This means that unlike the Impartial Spectator starting with a clean
slate – any allocation in the Edgeworth box is up for consideration – and advising Ayanda and Biko on the division of a pile of goods they have tripped
over in their student residence, market exchange starts from one particular
point in the Edgeworth box the endowment allocation.
The rules of the
E NDOWMENT ALLOCATION The ownership
of goods at the start of the game is termed
the endowment allocation. It is the starting
point of the game, but in applications to real
economies who owns what at any point in
time is the outcome (not the starting point) of
other interactions that have determine who
owns what.
P R O P E RT Y, P OW E R ,
B's coffee (kilograms), xB
8
7
6
5
4
3
2
Biko
1
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Pareto−efficient
curve
d
uA3 = 8.52
uA1 uAz
Pareto−improving
lens
tA
i
tB
uB3
uB4
uBz = 3.74 uB1
h
0
Ayanda
1
2
3
4
5
6
7
A
z
8
9
13
189
A's participation
constraint, uAz
12
11
10
tB
9
B's Utility, uB
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
9
B's data (gigabytes), yB
A's data (gigabytes), yA
10
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
8
7
i
6
Bargaining
set
5
4
B's participation
constraint, uBz
tA
z
3
2
Utility possibilities frontier
1
0
10
0
1
2
3
A's coffee (kilograms), x
(a) Edgeworth box for self-regarding Ayanda and Biko
4
5
6
7
A's Utility, uA
8
9
10
11
12
13
(b) The utility possibilities frontier
game then determine how the two can move to some other post exchange
allocation. The endowment allocation is important for two reasons:
• it is the starting point of the process and
• and also, because the exchange is voluntary meaning they can refuse to
trade, it is their fallback position, that is, the worst they can do.
Figure 4.7: Edgeworth box, the utility possibility
frontier, and the bargaining set. In Panel a uAz is
Ayanda’s utility at her endowment and is her participation constraint (shown by indifference curve
uAz ) and uBz is Biko’s utility at his endowment and is
his participation constraint (shown by indifference
curve uBz ).In Panel b) the coordinates of the x- and
y-axis intercepts of the utility possibilities frontier
give the utilities of the two when either Ayanda (
x-axis intercept) or Biko (y-axis intercept) have all
of the economic rents.
The participation constraint (PC)
To see how the second bullet above will narrow down that the post-exchange
allocation can be, starting at any given endowment allocation we introduce the
following notation, along with panel a. of Figure 4.7 (we will explain panel b.
below). The endowment bundle of person i is (xzi , yiz ) where the superscript
indicates who person i is (i = A for Ayanda, i = B for Biko). This allocation is
point z in the figure. It is identical to point z in previous figures, but instead of
being some hypothetical allocation that the Impartial Spectator was trying out
it is now something entirely different: it is what Ayanda and Biko own at the
start of the game.
From point z in the figure you can see that Ayanda’s and Biko’s endowments
of coffee and data are:
• Ayanda’s endowment: (xzA , yA
z ) = (9, 1)
• Biko’s endowment: (xzB , yB
z ) = (1, 14)
Introducing history in the form of initially privately owned endowments, along
with the voluntary transfer requirement, limits the possible allocations that can
result from exchange.
Because they can refuse any deal and therefore experience the utility from
R E M I N D E R The participation constraint is
also the fallback in the exchange scenario,
the utility that a person can certainly secure
if they choose not to participate in exchange
at all.
190
MICROECONOMICS
- DRAFT
their endowment bundle, they will not accept any post-exchange bundle that
makes them worse off than their fallback utilities. The indifference curves, uA
z
and uB
z , that include the endowment point are the post exchange bundles that
yield a utility identical to their fallback position.
These two indifference curves are called their participation constraints.
They are called participation constraints because Ayanda will not participate in
(that is she will refuse) any offer that would give her a post exchange bundle
below and to the left of uA
z . Likewise Biko will not participate in any offer that
would give him a post exchange bundle above and to the right of uB
z (labeled
as uB
z = 3.74 in Figure 4.7).
The right to refuse an exchange that will make a player worse off – the basis
of the participation constraints – reduces the possible set of post exchange
allocations consistent with voluntary exchange, and starting from the endowment allocation indicated by point z in panel a.
The yellow colored space between the two constraints – the indifference
B
curves, uA
z and uz – is the entire set of allocations that are Pareto superior
to point z and which therefore could be the result voluntary modifications
of the endowment allocation by means of exchange. This area is called the
Pareto-improving lens.
4.6 Symmetrical exchange: Trading into the Pareto-improving lens
PARETO IMPROVING LENSThe set of allocations that are Pareto superior to the fallback
options of the players is the Pareto improving
lens – shaded yellow in the figures to follow
in the rest of the book.
In this section we start with the assumption that the two traders have identical
preferences. That is, that their Cob-Douglas utility functions have a A = a B =
0.5.
We used a hypothetical point z in Figure 4.5 to show that an allocation where
the indifference curves cross cannot be Pareto-efficient. Our demonstration
consisted of showing that at such an allocation both Ayanda and Biko could
benefit from exchange.
We can now use the same reasoning to illustrate how starting at point z, now
an endowment allocation – a real distribution of ownership of two bundles –
the two could trade into the Pareto improving space, and eventually all the
way to the Pareto-efficient curve. Each person has a willingness to pay for x
in terms of y, their marginal rate of substitution at the endowment allocation
z. Ayanda’s maximum willingness to pay is her mrsA (9, 1) = 19 and Biko’s
maximum willingness to pay is his mrsB (1, 14) = 14.
The difference between Ayanda and Biko’s willingness to pay (mrs) signals an
opportunity for Ayanda to trade data with Biko at a rate of exchange between
her own marginal rate of substitution and Biko’s marginal rate of substitution.
A small exchange on these terms would move them to a post-exchange
allocation upward and to the left of the endowment.
With a A = a b = 0.5, their
marginal rates of substitution are mrsA =
M-CHECK
yA
xA
1
9 and
mrsB =
yB
xB
= 14. This means
they could make an exchange at a “price”
between 19 of a gigabyte for a kilogram of
coffee (Ayanda’s mrs) and 14 gigabytes of
data for a kilogram of coffee (Biko’s mrs).
=
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
To stress that the game is entirely symmetrical imagine that they have agreed
on a set of rules to determine the price and the amounts to be exchanged. At
any allocation at which the mrs of the two differs (meaning their indifference
curves intersect), take the following steps:
1. Pick a "price" midway between the mrs of the two. (This means that at
point z the price would be 14 + 19 divided by 2, or 7.06.)
2. Ask the amounts that each would like to transact at the price of 7.06 gb of
data for a kilo of coffee, for example how much coffee Ayanda would like
to ’sell’ at this price, and how much coffee Biko would want to ’buy’(these
desired amounts will differ between the two);
3. Because the transfer has to be voluntary (nobody can be forced to buy
more than they wish), transfer the amounts desired by the person who
wishes to transact least.
4. At the resulting post-exchange allocation determine if the indifference
curves are intersecting. If so return to step one and continue. If not (that is,
if the indifference curves are tangent) end the game with this final allocation.
We can see that by this process the two will have moved, step-by-step from
the endowment allocation at point z to a final post-exchange allocation that
will be on the Pareto-efficient curve. We know that they will get there for two
reasons:
• Trades are Pareto-improving: each trade they take moves them in the
direction of the Pareto-efficient curve because moving in the other direction could not be a Pareto-improvement and would violate the voluntary
transaction requirement.
• Trade concludes at a Pareto-efficient outcome: by the rules of the game
they have adopted they will keep on exchanging until they are at a place
where their mrs’s are identical, which must be on the Pareto-efficient curve.
They could have adopted a different set of rules or institution for exchange.
For example, they could have said that for step 1 above there will be two
alternative prices, one just a little less than Biko’s willingess to pay, and the
other just a little more than the lowest price at which Ayanda would part with
her coffee; and then just flipped a coin to see which of these prices they would
use in that transaction. Having made that transaction, do another coin flip to
see whose preferred price will be used, and so on until they reached a Pareto
efficient point at which no further trade was possible.
Other than knowing that they would eventually get to the Pareto-efficient
curve, we do not know which specific point on the curve they would get to. If
the coin flips went in favor of Ayanda, they could end up close to tA with Biko
191
192
MICROECONOMICS
- DRAFT
sharing very little of the gains from exchange. Or it could have gone the other
way, somewhere near point tB . They even could have ended up at point i the
allocation chosen by the Impartial Spectator.
The utility possibilities frontier in Panel b. of Figure 4.7 translates these allocations and the transactions supporting them into the utilities of the two players.
The Pareto-improving lens in panel a. corresponds to the bargaining set in
panel b. The first panels shows all of the allocations – denominated in quantities of x and y allocated to the two – that are Pareto improvements over the
endowment allocation. The second – the bargaining set – shows the utility
levels associated with every allocation in the Pareto-improving set.
The yellow area in panel b. is called the bargaining set because it compares
outcomes of their bargains relative to their fallback position (point z). The
bargaining set shows all of the possible distributions of the rents (utility greater
than their fallback positions) that might result from their bargaining, depending
on the rules governing how they bargain. These rules will determine the
extent of the bargaining power of the players.
Checkpoint 4.4: Pareto improvements, rents, and Pareto efficiency
If point h is the post exchange allocation based on the endowment allocation of
point z, explain the following:
a. Did Biko benefit from the exchange?
b. Did Ayanda benefit from the exchange?
c. What is the rent that Ayanda receives as a result of this exchange?
d. Did the exchange result in a Pareto improvement?
e. Is the post exchange allocation (point h) Pareto efficient?
4.7 Bargaining power: Take-it-or-leave-it
In the bargaining over the distribution of coffee and data above, the two examples of rules of the game were symmetrical. Neither "split the difference
between the willingness to pay of the two" or "alternating coin flips to see
whose preferred price will be used" gave any obvious advantage to either
player
But many bargaining interactions are asymmetrical. One of the players has
most of the bargaining power. Bargaining power is the ability to gain a large
share of the mutual gains from exchange (total rents) made possible from
some interaction, as may be determined by the rules of the game governing
the interaction and the skill of the players in securing a favorable agreement
under these rules.
An example is the Ultimatum Game in Chapter 2 (whose name already sug-
B ARGAINING POWER is the ability to gain
a large share of the mutual gains from
exchange (total rents) made possible from
some interaction, as may be determined
by the rules of the game governing the
interaction and the skill of the players in
securing a favorable agreement under these
rules.
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
193
gests the asymmetry). The Proposer makes a offer of some fraction of the
"pie". The Responder’s strategy set is simply: accept or reject, or "take it or
leave it." Being in a position to make that kind of an ultimatum is called take it
or leave it power, or TIOLI power for short.
In the coffee for data bargaining game if Ayanda had TIOLI power, she could
have said to Biko: “I’ll give you 2 kilograms of coffee and you give me 9 gigabytes of data. If you refuse, I will not agree to any other trade you might
TAKE - IT- OR - LEAVE - IT- POWER A player with
TIOLI power in a two-person bargaining
game can specify the entire terms of the
exchange – for example, both the quantity to
be exchanged and the price – in an offer, to
which the other player responds by accepting
or rejecting.
propose.” In other words, ”either accept the allocation I impose, or we both
stay at our endowment, z." The assumption that Ayanda’s offer is credible is
important: if Biko suspects that he could refuse and Ayanda would listen to a
counter-offer, the threat in the TIOLI offer would be empty.
A bargainer with TIOLI power can often capture most or even all of the total
rents that an economic interaction provides. This is because TIOLI power
allows a bargainer to specify both:
• the price at which the goods will be exchanged and
• the amount of goods that will be exchanged
This means that the person with TIOLI power can just pick some preferred
allocation – a point in the Edgeworth box different from the endowment point –
and make that the TIOLI offer.
What take-it-or-leave-it offer will Ayanda make to Biko?
Ayanda does not care about Biko’s utility, but she does care about how he will
respond to her offer. If he rejects, then she gets her fallback option. She will
realize that she must offer Biko a deal that Biko regards as better – or at least
not worse – than the endowment. In other words, Ayanda has to take Biko’s
participation constraint as a limit on the kind of offer she will make. This is
an example of the backward induction method that you learned in Chapter 2:
Ayanda has to reason backwards from her understanding of what Biko will do
after she has made her offer to what offer she should make now.
So Ayanda has the following constrained maximization problem: find a final
allocation (different from the endowments) to propose at which Biko is no
worse off than at the endowment and Ayanda is as well off as she can be.
Ayanda knows that the solution to this problem must have two characteristics:
It must:
• satisfy Biko’s participation constraint, that is, be in (or on the boundary of)
the Pareto-improving lens in Figure 4.7.
• be Pareto-efficient, but this is not because Ayanda cares any more about
efficiency than she does about Biko: if she offered an allocation that satisfied Biko’s participation constraint and was not Pareto efficient then there
R E M I N D E R The Ultimatum Game discussed
in Chapter 2 has this TIOLI structure including returning to the endowment point
the the Responder rejects – both getting a
payoff of zero, namely what they would have
received had they not interacted. That is
why it is called the Ultimatum Game, as the
Proposer’s offer is an ultimatum.
194
MICROECONOMICS
- DRAFT
would be some other allocation at which she could be better off and Biko
not worse off.
Ayanda would probably offer Biko something just a tiny bit better than Biko’s
fallback utility to make sure he accepts. But to avoid having to keep track
of that tiny amount in our thinking, here and in the rest of the book, we will
assume that Biko will accept an allocation that just meets his participation
constraint.
That solves the problem for Ayanda: to meet the two requirements bulleted
above , she must find the intersection of the Pareto-efficient curve and Biko’s
participation constraint uB
z . Therefore, Ayanda offers an exchange that imple-
M - C H E C K Remember that in Chapter 3,
a utility maximizer is often constrained by
a feasible frontier. Even with TIOLI power,
Ayanda is constrained by Biko’s participation
constraint, that is, uB (xB , yB ) uB
z.
ments point tA at the extreme end of the Pareto-efficient curve in Figure 4.7 a.
The same result is shown in Figure 4.7 b, where tA represents the distribution
of utilities resulting from the TIOLI allocation that Ayanda offered and Biko
(barely and grudgingly) accepted. Point tB at the other extreme of the Paretoefficient curve in the figure in the Edgeworth box corresponds to the allocation
where Biko has TIOLI power and point tB on the utility possibilities frontier is
the corresponding distribution of utilities.
We can see that the TIOLI allocation does not weight the two utilities identically (as did the social welfare function of the Impartial Spectator, which led to
point i). This why we say that allocation tA is Pareto-efficient but not socially
efficient, where the latter term is whatever the Impartial Spectator selected
based on maximizing an equally-weighted social welfare function.
Two features of the TIOLI outcome where the participation constraint holds
are important because they arise in many social coordination problems that
involve a participation constraint:
1. Pareto efficiency: The PC-constrained outcome is Pareto-efficient.
2. Inequality : At PC-constrained outcomes the bargainer with TIOLI power
gets all of the economic rent.
M-Note 4.4: Finding Ayanda’s TIOLI Offer
We need two pieces of information to find Ayanda’s TIOLI offer:
• The equation for the Pareto efficient curve (because we know that the resulting allocation will be Pareto efficient) and
• The equation for Biko’s participation constraint (because we know that Ayanda will not
offer him anything better than his utility at his endowment bundle).
The Pareto efficient curve: At Checkpoint 4.3 we asked you to find the Pareto-efficient
curve for Ayanda and Biko when they have identical Cobb-Douglas utility functions with
a = 0.5. The solution is that the Edgeworth box has the following Pareto-efficient curve
R E M I N D E R An outcome is socially efficient
when it maximizes a social welfare function;
what is deemed socially efficient depends
on how the utility of each member of the
population is weighted in the social welfare
function.
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
defined over the two people’s allocations of x and y:
yA
✓ ◆
3 A
x
2
=
(4.3)
(4.4)
We can re-write Equation 4.5 in terms of xB and yB by substituting xA
ȳ yB in the equation to find:
yB
= x̄
xB and yA
✓ ◆
3 B
x
2
=
=
(4.5)
As you can see, the Pareto-efficient curve is a line from the one corner of the Edgeworth
Box to the other. The utility functions, endowments and TIOLI offers calculated in this
M-Note are the basis for Figure 4.7.
To find the TIOLI offers, we need the players’ participation constraints because the player
with TIOLI power wishes to maximize their utility subject to the participation constraint of
the other player.
Biko’s participation constraint:
B’s fallback utility (his participation constraint (PC)) at his endowment xzB = 1, yB
z = 14 is:
uBz (1, 14)
=
(1)0.5 (14)0.5 = 3.74,
So we need to find the point on the PEC at which Biko has this level of utility.
A’s TIOLI Offer: We substitute the Pareto-efficient curve’s value for xA into B’s utility
function that equal to B’s fallback utility:
uB = (xB )0.5
✓
◆
3 B 0.5
x
2
| {z }
=
uBz = 3.74
| {z }
PC
PEC
3
) ( )0.5 xB
2
=
B
xTA
=
) yBTA
=
A
) xTA
A
) yTA
=
3
3.74/( )0.5 = 3.05 ⇡ 3
2
3 B 3
9
x = (3) = = 4.5
2
2
2
x̄ xB = 7
=
ȳ
3.74
yB = 10.5
So where “TA” means A had TIOLI power, the post-exchange allocation will be
A = 7, yA = 11.5, xB = 3, yB = 4.5 . The post-exchange allocations imply that A
xTA
TA
TA
TA
B
made a TIOLI offer to B of 2 units of x (xTA
y (yATA
yAz = 14
xzB = 3
1 = 2) in exchange for 9.5 units of
4.5 = 9.5). A’s utility is uATA = 8.97 and B remains on his participation
constraint at uB
z = 3.74.
Checkpoint 4.5: Biko’s TIOLI offer to Ayanda
Given the TIOLI offer you just saw for Ayanda in M-Note 4.4, what would happen
if the players’ positions were reversed and Biko had TIOLI power over Ayanda?
To answer this here is A’s fallback utility at her endowment xzA = 9, yA
z = 1:
uAz (9, 1)
a. What offer would Biko make?
=
(9)0.5 (1)0.5 = 3,
195
196
MICROECONOMICS
- DRAFT
b. What would the post-exchange allocations be? Explain.
c. Use Ayanda’s utility at her fallback position that we found in M-Note 4.4.
4.8 Application: Bargaining over wages and hours
We illustrate TIOLI power below by a case in which the two bargainers drop
their student personas to take on familiar roles in what is arguably the most
important market in a modern economy: Ayanda is the owner of a company
interacting with Biko, a prospective employee. In labor market bargaining
over wages and working conditions the employer almost always has TIOLI
power, stating the wage, the job and the hours. The worker accepts or not. We
postpone until Chapter 15 the question: why might Ayanda get to have this
power and not Biko?
So, leaving the world of coffee and data behind us, we will see that the sum of
the mutual gains enjoyed by the two and how these are divided between them
will depend on both their preferences and the rules of the game:
• Power: Do the two bargain symmetrically with neither one nor the other of
them having first mover advantage? Is one of them first mover with take-itor-leave-it power (TIOLI power)?
• Fallback: What is each person’s fallback position? how well off are they
if they do not exchange at all? Does Biko have other options than being
employed by Ayanda? If Ayanda does not employ Biko, are there others
she she could employ?
To fill in some answers to those questions, our two actors are now:
• Ayanda, an employer : whose endowment bundle is a sum of money only
(no employees), and who in the absence of any exchange with Biko has
nobody work for her; she will make Biko a take it or leave it offer of a sum
of money in return for some number of hours of work for her, and
• Biko, a worker : who is applying to work in Ayanda’s company, whose
endowment bundle is free time only (no money); he has a maximum of 16
hours of (non-sleeping) time to spend, possibly working for Ayanda.
We introduce a more complete model of the labor market with competition
among firms for workers and customers and among workers for jobs in Chapter 11 including the ways that unemployment benefits, competition among
firms, and antitrust could affect these outcomes.
Quasi-linear preferences for money and time
To represent the preferences of Ayanda and Biko we will introduce a new
form of utility function, one that will simplify our analysis while still conveying
Q UASI - LINEAR FUNCTION A quasi-linear
function depends linearly on one variable,
e.g. y, and non-linearly on another variable,
e.g. x, and has the form u(x, y) = y + h(x),
where h(x) is a non linear function.
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
Figure 4.8: Marginal rates of substitution with
quasi-linear preferences. With quasi-linear
preferences, marginal rates of substitution depend
only on the amount of the good x (here, Hours of
Living for Biko), and not at all on the amount of
money left over to buy other goods, y. As a result,
indifference curves with different levels of utility
are vertical displacements of a single curve –
you can add or subtract an amount of y from the
indifference curve to move it up or down.
y3 = 400
y2 = 340
Quantity of money, y
197
y1 = 260
h
g
e
uB3
uB2
f
uB1
0
0
2
4
6
8
10
Hours of Living, x
12
14
16
the main insights. The function is called quasi-linear because utility is partly
(“quasi") proportional to one of the arguments of the function, while being nonlinear in the other arguments. The Cobb-Douglas utility function is not quasi
linear because it is non linear with respect to both x and y.
As in the case of Harriet deciding how much fish to buy from Alfredo or Bob
in Chapter 3, we will consider the second good as “money left over" after the
exchange. This may seem odd because money is not something you value for
itself. But money can buy you other goods which you do value: The utility of
"money left over" is the utility of the goods which the person can purchase as
a result.
We now illustrate a case where one person starts off with all of one good and
none of the second, so the other person all of the second good,but none of
the first. This could model you walking into the supermarket with money in
your pocket (or more likely a credit card) and nothing in your shopping bags,
and planning to walk out with less in your credit card and some groceries in
our shopping bag. So it is a model of any kind of exchange. But here illustrate
it by Ayanda (possibly) employing Biko.
The marginal rate of substitution for a person with quasi-linear preferences
that are linear in "money" (y) depends only on the amount she has of the
good or service for which her preferences are non-linear, not on the amount of
money.
The reason why this is true is because:
198
MICROECONOMICS
- DRAFT
• The marginal rate of substitution is the ratio of the marginal utility of x to the
marginal utility of y
• The person’s marginal utility for y is always a constant and it does not
decline as she gets more y, so
• The marginal rate of substitution depends only on the marginal utility of x
which varies with the quantity of x consumed because the function is non
linear in this variable.
You can see this in Figure 4.8 by noticing that for a given amount of x the
slope of the indifference curve (shown by the dashed tangent lines) is the
same no matter how much y the person has, such as at x = 8 hours of living,
as shown by points f, g, and h. This is because, given the quasi-linear utility
function that we used to draw the figure, the willingness to pay for an additional hour of living (the marginal rate of substitution, that is, the negative of
the slope of the indifference curve) does not depend on the amount of money
lef over that the person has; it depends only on how many hours of living they
have. This means that the indifference curves u1 , u2 , and u3 in the figure are
just shifted up replicas (you can see the amounts by which they are shifted up
by comparing the vertical axis intercepts).
We can also compare points e and g at the same level of y: Biko likes to have
B
more living time (uB
1 < u2 ) and his willingness to pay for additional hours of
Living declines the more he has (the indifference curve is less steep at g than
at e).
It is of course unrealistic to think that anyone would have truly linear preferences in any amount imaginable of money left over, for this would require that
the person did not have diminishing marginal utility in the things that money
can buy. But because "money" can be considered as generalized purchasing
power that can be spent on a vast array of things, and because we do not consider changes in people’s bundles of money making them either billionaires or
paupers, it is a useful simplifying assumption.
M-Note 4.5: A quasi-linear utility function
Quasi-linear utility functions have the form:
u(x, y) = y + h(x)
(4.6)
Equation 7.29 is quasi-linear because it is linear in one variable, y, and non-linear in the
other x as in the non linear function h(x). For example, h(x) could be quadratic in x or
could include the natural log of x, ln(x).
A quasi-linear utility function depends linearly on one variable, e.g. y, and non-linearly
on another variable, e.g. x, and has the form u(x, y) = ay + h(x), where a is a constant.
Hence it is quasi or ‘partly’ linear. We often set a = 1. Two examples of such functions
include, u(x, y) = y + 20x x2 and u(x, y) = y + 10 ln(1 + x).
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
199
Allocating money and time
For simplicity, we assume that Ayanda, the employer, and Biko, the worker,
have quasi-linear utility functions. Both place a constant value per monetary
unit on the money they have after the transaction (to purchase other things,
for example). That is, the marginal utility of money is constant. So we can
measure the utility of each in whatever monetary units they are using, which
since their names are from South Africa, might as well be the South African
Rand.
Biko values his Living (that is his 16 waking hours, minus the time he "hires
out" of himself to work for Ayanda). But the marginal utility of free time is
decreases as the amount of free time he has increases, just as was the case
for Aisha in Chapter 3. His first hour of free time is high, but his free time gets
less and less valuable the more free time he has.
Ayanda places a value, too, on Biko’s free time, but it is the opposite of Biko’s
value: she benefits by Biko having less free time and her having more of
Biko’s time working for her. The positive value she places on Biko’s labor –
like the positive value he places on his free time – depends on how much of it
she gets. The marginal utility to Ayanda of Biko’s labor decreases as she hires
more of his time: the value of Biko’s work is high the first hour Ayanda hires,
less valuable the second hour, less valuable the third, and so on.
This is because if she has just an hour of his time, she assigns him to really
important tasks, but the tasks he does in later hours are less essential to
Ayanda. (This is similar to why the marginal productivity of time studying
diminishes as the amount of time studying increases). Not accounting for the
wage she must pay him, each hour of Biko’s work gives Ayanda utility, but at a
decreasing rate.
Figure 4.9 shows the setting for this interaction as an Edgeworth box, with
the quantities interpreted amounts per day. The endowment point z is in the
upper left corner of the box showing that Biko has 16 hours of Living time and
no money. Ayanda has $400 but no Labor from Biko to work in her company.
As before, like z every point in the box represents an allocation that is feasible
given the amount of money that Ayanda has in her endowment bundle and the
amount of free time that Biko has in his.
Three of Biko’s indifference curves and three of Ayanda’s are shown in Figure
4.9. For both Ayanda and Biko, their reservation indifference curve (their
participation constraint) goes through the endowment point where Biko has
16 hours of Living and Ayanda has $400 per day to pay workers, meaning that
the indifference curves include the endowment point z.
Also shown is one of Biko’s indifference curves labeled uB
3 , which is tangent to
B
Ayanda’s participation constraint (uA
z ) at point t . The allocation given by the
F AC T C H E C K At the time of writing this 1
Euro was equal in value to about 16 South
African Rand (ZAR), 1 Pound Sterling was
equal to about 21 South African Rand. In
2020, the hourly minimum wage in South
Africa was ZAR 20.76.
MICROECONOMICS
14
12
B's Hours of Living,xB
10
8
6
4
2
Biko
0
0
16
- DRAFT
z
uBz = 256
B's PC
tA
uB3 = 508
100
uA2 = 516
tB
300
200
j
200
Pareto−efficient
curve
uB2 = 392
A's PC
uAz = 400
0
100
B's Money, yB
A's Money, yA
300
uA3 = 652
0
Ayanda
2
4
6
8
10
12
A's Hours hired of B's Work, xA
Figure 4.9: Bargaining over hours and wages.
Shown are three each of Ayanda’s and Biko’s
indifference curves and the utility that they
experience at any of the allocations indicated
by the points making up these curves. Point z
is the endowment allocation which is a point on
the participation constraints of each of the two.
Points tA and tB respectively are the allocations
resulting when Ayanda or Biko are first mover
with TIOLI power. The yellow shaded area is the
Pareto-improving lens. The vertical line (including
its dashed portions) is the Pareto-efficient curve
made up of all points of tangency between the
indifference curves of the two such as j, tA and tB .
14
16
400
400
200
that tangency is a Pareto-efficient allocation (because the marginal rates of
substitution of the two are equal). We also show a third indifference curve for
B
Ayanda, labeled uA
3 , which is tangent to Biko’s participation constraint (uz ) at
point tA . These two tangencies are points on the Pareto-efficient curve, which
is a vertical line through these points all the potential tangencies above each
person’s fallback.
The reason why the Pareto-efficient curve is vertical here (remember it was a
diagonal line in the previous Edgeworth boxes) is that Ayanda and Biko have
quasi-linear utility functions. With quasi-linear utility, the marginal utility of
hours depends only on the quantity of hours and not on the amount of money
they have. If the two curves are tangent at 8 hours when Ayanda has most of
the money and Biko little, they will also be tangent at 8 hours when Biko has
most of the money and Ayanda has little.
4.9 Application. The rules of the game determine hours and wages
The Edgeworth box and the indifference curves by themselves do not determine the outcome of the interaction. Without knowing more, any point in
the box is a possible outcome. Knowing the endowment allocation z narrows
down the possible post exchange allocations but not by very much.
Employment in most modern economies is voluntary (but see the Fact
Check), so we will assume that the outcomes are limited to those that are
at least as good for each participant as their fallback position given by point
z. As a result, outcomes of bargaining between the employer and the worker
must be in the yellow shaded Pareto-improving lens in Figure 4.9.
F AC T C H E C K In the past slavery has
meant the ownership of one person by
another, including the right of sale of the
slave to another owner. The term modern
slavery refers to any situation in which, like
historical slavery, the services or goods
that one party provides for another are not
voluntarily offered but are motivated by fear
of severe harm. Ownership of one person by
another need not be part of modern slavery.
Prisoners, immigrants without legal rights
of residence, residents of undemocratic
countries, “sex slaves," and children are over
represented among contemporary ’modern
slaves.’
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
201
We illustrate the importance of institutions by showing the allocations will
result under four different rules of the game. Each set of rules is a specific account four different ways that an employer and worker might interact. :
• The employer can make a take-it-or-leave-it offer of both the wage and the
hours worked.
• As members of a trade union, the employees (we will take Biko as a representative worker) can make a take it or leave it offer specifying both the
wage rate and the length of the working day (hours).
• Legislation is passed limiting working hours per day to no more than 5
hours and the total pay or this period to not less than 254 Rand or 50.80
Rand per hour.
• The above legislation is passed, but it has a proviso that if the two parties can agree on an alternative allocation their agreement can be implemented.
Employer has TIOLI power
Imagine that, like most employers, Ayanda can offer Biko a job description:
work a given amount of hours for a given amount of pay (and therefore for a
particular hourly wage). Biko’s only choice is to accept or reject, so Ayanda
has take-it-or-leave-it power. For Biko to accept, Ayanda knows the offer must
be at least as good as Biko’s reservation option, so the relevant constraint for
her is Biko’s participation constraint (as was the case for the coffee and data
bargaining).
She will choose the point that she values most along this indifference curve,
and therefore implement an offer indicated by point tA . Having TIOLI power,
the employer has gotten all of the economic rent, leaving Biko indifferent
between taking the job and refusing it (as before in cases like this we just
assume he takes the job).
What is Ayanda’s rent from this transaction, meaning the excess of her utility
at point tA compared to at point z, the endowment allocation at which no trade
has occurred? Reading the utility numbers from her indifference curve at point
tA and her reservation indifference curve through point z we can see that her
A
rent is uA
3 = 652 minus uz = 400 or 252. Because utility is measured vertically
in terms of money this is the same thing as the vertical distance between
points tA and tB in the graph.
Employees and their trade union have TIOLI power
Turning to the opposite case Biko, through his trade union, is now first-mover
with TIOLI power. The offer he will make (and she will accept) is the opposite
E X A M P L E Put yourself in Biko’s shoes if the
allocation is point tA . How do you think he
feels about his employer and his job? Would
he be motivated to work hard, not to steal
from his employer, and otherwise contribute
to the profitable operation of her firm? These
are serious problems and a reason why
extreme allocations – like Ayanda getting all
of the economic rent from the interaction and
Biko being indifferent between his job and
being fired – are not commonly observed.
If Ayanda has an interest in Biko’s good will
and hard work, she may have to share at
least a bit of the gains from exchange with
Biko so that he receives a rent. This fact will
become important when we consider the
labor market.
MICROECONOMICS
12
B's Hours of Living,xB
8
Biko
4
0
0
16
- DRAFT
z
100
uA3 = 652
Pareto−
efficient
curve
a
200
A's Money, yA
b
B's Money, yB
uB2 = 351
tA
200
300
uBz = 256
100
uB3 = 508
300
uA2 = 540
tB
decreased
work
hours
0
uAz = 400
0
Ayanda
4
8
12
A's Hours hired of B's Work, xA
16
400
400
202
of tA the allocation resulting when Ayanda had TIOLI power. Biko will recognize Ayanda’s participation constraint – he as to make her an offer she will
not refuse. And he will choose the allocation indicated by point tB in which his
post exchange bundle gives him all of the economic rents of 252.
This is the most that Biko could demand without Ayanda simply going out
of business. This constraint on the demands that workers can make on employers in a market and profit based economy will be a major theme in the
chapters to come.
Legislation imposes hours and pay limitations.
As described above the legislation imposes on both Ayanda and Biko the
allocation at point b. But the allocation imposed by the legislation is Pareto
inefficient. But it does set a new status quo, a fallback position that, if they
cannot come to some agreement will be the post exchange allocation.
Both Ayanda and Biko can see that at b they could both do better by agreeing that Biko should work more than 5 hours, and Ayanda should pay him
more. The yellow Pareto-improving lens shows the space for their possible
bargains.
Bargaining to override the legislation: more work and more pay.
They could bargain to agree upon any point in the yellow Pareto improving
lens, possibly agreeing on the Pareto efficient allocation at point a. Where
they ended up in or on the boundary of the Pareto improving lens would
depend on the rules of the game governing that bargaining process. They
might even fail to agree on any bargain – as is often the case with players in
Figure 4.10: Allocations with legislation and
bargaining. The legislation stipulating hours and
pay results in the allocation indicated by point b.
Because b is preferred to the no exchange option
z by both of them, they will definitely make an
exchange. But then can both do better than at
b. Taking the allocation at b as their new fallback
position, they could bargain to point a or any other
allocation in the yellow Pareto improving lens.
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
the Ultimatum Game – and remain at point b.
Figure 4.11 shows how Ayanda and Biko do under these differing rules of the
game as indicated by the rents they enjoy in the Nash equilibrium of each
game, that is the utility associated with their fallback options 400 and 256
respectively.
Introducing a historically realistic set of rules of the game – making the employer the first mover with TIOLI power – has two effects: it generates 252
units of utility in gains from trade, and it makes the final allocation more unequal than the endowment allocation (because the employer captures all of
the mutual gains made possible by exchange). Biko’s share of the total utility
(not shown) falls from two fifths to less than a third.
In many countries during the 20th century the response to the unequal allocations implemented when the employer has TIOLI power was the formation
of trade unions. And you can see from the figure that if the union is powerful
enough for it to have TIIOLI power, then Biko (and his trade union colleagues)
capture the entire rent, Ayanda getting nothing more than her reservation
utility. Biko’s share of the total utility jumps from two fifths at the endowment
allocation to well over half.
Even before workers had the right to vote and before trade unions were legal,
political movements mobilized to pressure governments to regulate working
conditions. In the model the introduction of hours and wage regulations implemented an outcome in which both Biko and Ayanda captured some of the
gains from trade. The reforms implemented a Pareto inefficient allocation but
the shortfall from the maximum possible joint rents was minor (from 252 to
235).
The final case – bargaining up from the regulated hours and wages – describes labor markets in many countries today. Government regulations establish a fallback position, and then employers and workers (either individually or
in trade unions) seek bargains that improve on that allocation.
Though they differ radically in their distributional aspect, all of the scenarios are Pareto superior to the endowment allocation. We can also see that
the negotiated allocation after legislation is Pareto superior to the allocation
implemented by the legislation.
We cannot say which of the three Pareto efficient allocations is preferred from
a fairness standpoint without knowing more about Ayanda and Biko’s other
wealth, their needs, and other aspects that might affect their ethical claims on
the benefits of their interaction.
Checkpoint 4.6: Bargaining over hours and wages
203
204
MICROECONOMICS
- DRAFT
Type
B's rents
A's rents
Gains from exchange
0
Employer (A) has TIOLI power, t
A
Union (B) has TIOLI power, t
B
252
252
252
0
Figure 4.11: Rents under differing rules of the
game, with Ayanda as employer and Biko as
worker The rents and gains from exchange of
each set of rules are shown in the figure. That is,
the figure shows each player’s utility under each
set of rules minus that player’s fallback option
(uAz = 400 and uBz = 256 respectively). The gains
from exchange are the sum of the rents received
by Ayanda and Biko. Source: Authors calculations
described in the text.
252
95
Legislated hours and wages, b
140
235
104
Negotiated allocation after legislation, a
148
252
0
100
Rents
200
300
Using the figure, explain how the following two things (taken separately) would
affect the outcome under the four different rules of the game above (start by
explaining how the endowment point z would be affected):
• If Biko does not exchange their time with Ayanda and is unemployed, he
receives what is called an unemployment benefit, that is a payment from the
government equal to $100, and this is financed by a tax on Ayanda equal to
$100.
• Ayanda now has free access to a robot that will at no cost do work equivalent
to two hours of Biko’s time.
4.10
First-mover advantage: Price-setting power
Returning to Ayanda and Biko with their former personas as students exchanging coffee and data, we will now see that while first movers typically
have advantages, these advantages may not be due to TIOLI power. Ayanda
may be first-mover but be unable to commit to take-it-or-leave-it offer that
stipulates an exchange of a specific amount of coffee for a specific amount of
data.
Price-setting power
She may have what is called price-setting power (PS power) if she can specify a price – either a monetary price or the ratio at which the two will exchange
goods – but not how much (the quantity) of her good Biko will buy.
Ayanda might say, for example: “I will give you one kilogram of coffee for
every three gigabytes of data you give me. You can decide how much data
F IRST- MOVER ADVANTAGE A player has a
first-mover advantage when the institutions,
history, or power structures of a game give
the player the opportunity to make an offer or
move before the other players in the game
can take action. The opportunity to move
first can confer an advantage that results in
higher utility or a greater share of economic
rents in the outcome of an interaction.
P RICE - SETTING POWER A first-mover with
price-setting power (PS power), can commit
to a price – the ratio in which goods will be
exchanged – but not the quantity that will be
transacted at that price.
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
205
you would like to exchange for coffee at that ratio, but the ratio is not going to
change. Of course you are free to buy nothing.”
We saw that owners of companies typically have TIOLI power when hiring
employees; but in their interactions with their customers they typically have
price-setting power. They set a price at which they will sell their product,
and sell as much to each customer at that price as the customer wants to
buy.
The incentive compatibility constraint (ICC).
If Ayanda has price setting power she must find a way to determine the price
when it is the price alone that makes up her offer. So her constrained optimization problem is not the same as it was when she had TIOLI power.
When Ayanda had TIOLI power she had only to satisfy Biko’s participation
constraint: if her take-it-or-leave-it offer were a post exchange bundle that
would make Biko worse off than at his endowment bundle, Biko would leave
it rather than taking it! Of course whether she has TIOLI power or just price
setting power, if Ayanda wants to exchange with Biko, she will have to satisfy
his participation constraint.
But there is now a second constraint she must satisfy called the incentive
compatibility constraint. What this means is whatever post-exchange bundle Ayanda would like to implement, she must provide Biko with incentives
so that his best response will be to exchange the amount that will allow her
to "move" from her endowment bundle to her desired post-exchange bundle.
This is called the incentive compatibility constraint because she must provide
Biko with incentives that motivate Biko to act in a way that is compatible with
(meaning, that implements) her desired outcome. The incentive compatibility
constraint is based on Biko’s best response – the amount of coffee he is
willing to buy – to the price Ayanda offers.
You have encountered best responses in Chapter 1. There the the strategy
sets were particular actions and therefore best responses were limited to
actions like “Plant Late," or "Fish 12 hours." Options like "Plant a little earlier"
or "Fish 10 hours and 15 minutes" were not possible. Sometimes discrete
strategy sets and best responses like this make sense (think: "Drive on the the
left if you are in the U.K., or Japan").
But sometimes the strategy sets for players are continuous, as for example in
setting a price for a good or when choosing the amount of time for an activity,
like fishing. When this is the case – as with Ayanda’s decision to set a price –
we consider the players’ best responses as continuous variables and describe
them by best-response functions.
M - C H E C K A continuous variable can take
on any value over some interval. So, a
variable that can take the value of any
number between 0 and 5 is a continuous
variable; a variable that is restricted to the
integers between 0 and 5, namely, 1, 2, 3 or
4 is discrete. The number of your sisters or
brothers is discrete, the height any one of
them is continuous.
I NCENTIVE C OMPATIBILITY C ONSTRAINT
The incentive compatibility constraint, ICC,
requires that first mover provide incentives
that make the second mover’s best response
be to act in ways that implement the post
exchange allocation which the first mover
prefers.
206
MICROECONOMICS
- DRAFT
As was the case when she had TIOLI power, Ayanda will reason backwards
from her understanding of what Biko how Biko will respond to each of her possible offers and how that will affect her utility. That is, she will use backward
induction.
To determine how Ayanda can maximize her utility subject to Biko’s incentive
compatibility constraint (the price-setting case) is a somewhat more complex
problem than maximizing her utility subject only to Biko’s participation constraint (the TIOLI power case). The reason is that in the TIOLI case there are
just two things that Biko can do: accept or reject her offer. But when Ayanda
has price-setting power only, Biko can choose from the entire range of possible amounts that he might be willing to exchange with her, depending on the
price.
As a result, Ayanda has to think in two stages when choosing a price ratio.
First stage: What will Biko do? How much coffee will Biko buy at each price
ratio Ayanda offers? This is Biko’s price-offer curve, which is his best
response.
Second stage: What should I do, given what he will do? Given her estimate
of Biko’s best response, which price ratio maximizes Ayanda’s utility? That
is, which price ratio takes Ayanda to her highest indifference curve, given
the constraint of Biko’s price-offer curve?
Best response and incentive compatibility
For the first stage, that is, determining how Biko will respond to each price
she might offer, Ayanda uses whatever information she might have, such as
her experience in the past with Biko’s response to offers, her best guess as
to Biko’s utility function, or her experience with other people she thinks are
similar to Biko.
Just as in Chapter 3 there is a budget constraint limiting the exchanges he
can undertake, but this is now a line giving feasible combinations of data and
coffee available to him through exchange at some given price.
If the price p – the number of gigabytes of data per kilogram of coffee – and
his post-exchange bundle is denoted as (xB , yB ) then Biko’s budget constraint
requires that the value of his post-exchange bundle must be the same as the
value of his endowment bundle, or:
pxB + yB = px̄B + ȳB
or p(x
B
B
x̄ ) =
B
(y
B
ȳ )
(4.7)
(4.8)
The second version of the budget constraint means that the value of the
coffee that he acquires (at the price p) or xB
x̄B must be equal to the value
R E M I N D E R The method is identical to
how we derived Keiko’s price-offer curve –
offering money in return for fish – in Chapter
3, except that here Biko is not ’buying’
coffee using money, he is exchanging data
for coffee. As a result the "price" is not in
terms of dollars per kilogram of coffee, but
gigabytes of data per kilogram of coffee.
P R O P E RT Y, P OW E R ,
of the data that he gives up ȳB
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
yB .
We can rearrange Biko’s budget constraint another way to show that the price
p must be equal to the ratio of the amount of data he gives up to the amount
of coffee he gets
p=
ȳB
xB
yB
x̄B
(4.9)
We show the derivation of Biko’s best-response function in Figure 4.12. We
start, in panel a by showing Biko’s best response to one particular price.
Then, in panel b we repeat the same reasoning for many prices, showing how
his best response to any price can be determined.
We know that given the price p4 Biko will choose how much data to transfer
to Ayanda in return for her coffee in order to maximize his utility subject to
his budget constraint. In panel a we show his feasible set with his budget
constraint for that particular price p4 its feasible frontier. The budget constraint
includes the point z because one of the feasible choices he could make while
respecting the budget constraint is to exchange nothing.
In Figure 4.12 the slope of the p4 line is the amount of data that Biko gives up
(DyB ) divided by the amount of coffee that he gets (DxB ) , when the price is
p. So:
p
=
DyB
ȳB
= B
B
Dx
x
yB
x̄B
marginal rate of transformation (mrt)
=
slope of the price line
For any given price this is the kind of individual utility maximization problem
that you studied in Chapter 3 in which the solution is to find the allocation at
which the mrs = mrt rule holds. You can see in panel a that the highest indifference curve that Biko can reach, consistent with his budget constraint (labeled uB
2 ) is tangent to his budget constraint at point b4 . This result expresses
the principle of constrained optimization that you have already learned. It is a
point equating:
• The slope of his indifference curve, which is the negative of the marginal
rate of substitution and
• The slope of the feasibility frontier – in this case the budget constraint –
which is the negative of the marginal rate of transformation of coffee into
data.
The mrt is the price p set by Ayanda, that tells Biko how many gb of data
he has to give up to get a kilo of coffee. Biko’s best response is to choose a
207
MICROECONOMICS
8
B's coffee (kilograms), xB
7
6
uBz , B's PC
5
4
3
2
1
Biko
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
p4
B's feasible set
uB2
Coffee
B gets
b4
Data B
gives up
z
0
Ayanda
1
2
3
4
5
6
7
A's coffee (kilograms), xA
8
9
10
10
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
9
8
p2
b2
4
3
2
1
Biko
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
uB3
b3
b4
uB4
B's best−
response
function
(ICC)
z
0
Ayanda
1
2
3
4
5
6
7
A's coffee (kilograms), xA
8
9
10
(b) B’s best-response function (ICC)
(4.10)
and, budget constraint: pxB + yB = px̄B + ȳB abbud
(4.11)
Equation 4.10 expresses the optimizing part of Biko’s choice, while Equation ?? expresses the constraint. The utility Biko enjoys at b4 in the figure is
B
greater than the utility of his endowment bundle (uB
2 > uz ). From this we conclude that if the price is p4 Biko will choose the post-exchange bundle given
by point b4 . This gives us one point on Biko’s best-response function.
In panel b we construct Biko’s best response function, by repeating the analysis in panel a but for differing prices tracing out a curve in the (x, y) coordinates. This is his best-response function because, by construction, points
on the curve show for each the value of p the post-exchange allocation that
maximizes his utility if could buy any amount of Ayanda’s coffee at the price p.
Ayanda now has all the information she needs to set the price.
M-Note 4.6: The incentive compatibility constraint
Here we show the derivation of the incentive compatibility constraint for Ayanda’s utilitiy
choice of a utility maximizing price to offer Biko. This equation will show, for every price
that Ayanda could offer, the amount of goods that Biko will be willing to exchange.
To do this we use Equations ?? and 4.10, the two conditions that Bikos response must
satisfy.
Given the price p offered by Ayanda, Biko’s budget constraint is
yB ( x B ) =
5
uB2
mrs = mrt tangency: mrsB (xB , yB ) = mrt = p
(4.12)
pxB + px̄B + ȳB
To maximize his utility uB (xB , yB (xB )), Biko will choose the bundle
6
p4
post-exchange bundle that satisfies the two conditions:
That is
7
p3
(a) B’s best response to a price = p4
pxB + yB = px̄B + ȳB
B's coffee (kilograms), xB
B's data (gigabytes), yB
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
9
B's data (gigabytes), yB
A's data (gigabytes), yA
10
- DRAFT
A's data (gigabytes), yA
208
(xB , yB (xB )) satisfies
Figure 4.12: Constructing B’s best-response
function (ICC). In panel a, B’s feasible set is in the
upper right corner of the Edgeworth box because,
as we explained in Figure 4.4, the upper left corner
of the box is the origin for him (indicating zero
of both goods). In panel a, when the price p4 is
equal to 3.53 Biko reaches his highest feasible
indifference curve (uB2 ) by giving up 5.3 gb of data
in return for 1.5 kg of coffee. In panel b he chooses
post-exchange bundles indicated by points b3 and
b2 in response to prices p3 < p4 and p2 < p3 .
B’s best-response function (ICC) connects these
and similar points all of them B’s utility-maximizing
bundle, for different prices.
P R O P E RT Y, P OW E R ,
the first-order condition,
uBx + uBy
dyB
= uBx
dxB
That is
1
uBy p = 0
uBx
=
uBy
mrsB (xB , yB ) =
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
p = mrt
(4.13)
2
Suppose that uB = (xB ) 3 (yB ) 3 , we can derive the incentive compatibility constraint using
Equations 4.12 and 4.13. From M-Note 4.2, we have
mrsB (xB , yB ) =
1 yB
2 xB
Moreover, the budget constraint can be rewritten as Equation 4.9, i.e.,
p=
Therefore, we have
1 yB
=
2 xB
ȳB
xB
yB
x̄B
ȳB
xB
yB
x̄B
(4.14)
which defines the incentive compatibility constraint shown in the Edgeworth box.
4.11
Setting the price subject to an incentive compatibility constraint
Biko’s best-response function is the incentive compatibility constraint for
Ayanda’s optimizing problem, shown in Figure 4.13. Notice that the incentive
compatibility constraint is more limiting to Ayanda than is Biko’s participation
0
constraint labeled uB
z , B sPC. This means that there are some allocations
(between the participation constraint and the incentive compatibility constraint)
which would make Biko better off than at his endowment bundle, and which
Ayanda would prefer to any point in her feasible set, but which Ayanda could
not implement when she has price-setting power but not take-it-or-leave-it
power.
Because Ayanda always has the option of simply discarding some of the data
she gets from Biko, we can think about the green shaded area under Biko’s
best response function as her feasible set. The slope of Biko’s best-response
function is (from Ayanda’s viewpoint) the marginal rate of transformation of
coffee into data, given how Biko responds to each of the prices she could
set. You can see that starting at the endowment allocation, the best-response
function is initially steep, so a modest amount of coffee that she gives up is
transformed – through exchange – into a substantial amount of data. But the
more data she wishes to acquire – moving up on the best-response function –
the less favorable to her the mrt becomes.
Ayanda’s choice of what price to set is a familiar constrained optimization
problem. It proceeds in two steps:
1. Determine the final allocation she would like to implement by finding the
point in the feasible set that is associated with the higher utility. To do this
she uses the mrs = mrt rule and selects point n in the figure, with its
associated utility uA
N (which exceeds that associated with point w, namely
209
MICROECONOMICS
8
B's coffee (kilograms), xB
7
6
5
4
3
2
Biko
1
0
uBz , B's PC
w
n
uAN
B's best−
response
function
(ICC)
A's feasible set
uAw
uAz
z
0
Ayanda
1
2
3
4
5
6
7
A's coffee (kilograms), xA
8
9
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
10
(a) Maximizing utility subject to the incentive compatibility constraint
10
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
9
B's coffee (kilograms), xB
8
7
6
5
uBz , B's PC
4
3
2
Biko
1
0
pN
uBN
tA
Pareto−
improving
lens
n
uAN
B's best−
response
function
(ICC)
uAz
z
0
Ayanda
1
2
3
4
5
6
7
A's coffee (kilograms), xA
8
9
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
B's data (gigabytes), yB
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
9
B's data (gigabytes), yB
A's data (gigabytes), yA
10
- DRAFT
A's data (gigabytes), yA
210
10
(b) Pareto-superior alternatives to Ayanda’s choice of n
uAw , which was also feasible. This is where her indifference curve is tangent
to Biko’s best best response function. This is shown in Panel a of Figure
4.13
2. Determine the price that will implement this outcome. Every allocation on
the best response function corresponds to some particular price that will
implement it. Price pN shown in Panel b of Figure 4.13 implements point n
We have given the price that Ayanda sets a superscript N because the allocation that it implements is a Nash equilibrium. To confirm that this is the case
we ask two questions:
• Given the strategy that Ayanda has adopted – that is, setting the price pN –
is there any way that Biko do better than he does by trading with her so as
to implement her chosen allocation (point n)? The answer is no, because
n is a point on his best response function, which tells us that if she offers
the price pN the best he can do is to trade with her so as to implement her
desired point.
• Given the strategy that Biko has adopted – his best-response function –
is there any way that Ayanda could do better than she does by setting the
price pN ? The answer is no, because she found point n exactly by doing
the best she could given his best-response function.
There are two important aspects of the Nash equilibrium (n) of this game.
First, the Nash equilibrium is not Pareto efficient. Ayanda’s and Biko’s indifference curves are not tangent at n, they intersect, and you know from the
mrsA = mrsB rule for a Pareto efficient outcome that any allocation at which
the indifference curves intersect is not Pareto efficient (because then the rule
is violated). The reason why Ayanda implemented an Pareto inefficient alloca-
Figure 4.13: A sets the price subject to B’s
best-response function (ICC) Ayanda’s utilitymaximizing post-exchange bundle is indicated by
point n where her indifference curve is tangent
to Biko’s best-response function (his price-offer
curve). The negative of the slope of the solid gray
line through both n and the endowment point z
is equal to the price Ayanda chooses, pN . Biko’s
budget constraint given by Ayanda’s choice of pN
is tangent to Biko’s indifference curve through n
by construction, that is, because n is on Biko’s
best-response function. To interpret the lower
shaded area as a feasible set, it must be the case
that A could choose not to consume the data or
coffee she has in that area (that is, some of it could
be thrown away).
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
tion is that the the constraint she faced was not Biko’s PC (the slope of which
is mrsB ) but instead his best response function (the slope of which is the
mrt ). So she implemented mrsA = mrt 6= mrsB violating the Pareto efficiency
rule. The allocations that are Pareto-superior to n are shown by the yellow
lens between the indifference curves through n.
Second, the person who is not the first mover (Biko) receives a rent in the
Nash equilibrium: as you can see from Figure 4.13 at n he is better off (on
a higher indifference curved) than with his endowment bundle (which is his
fallback option, namely no trade) indicated by the indifference curve labeled
uBz , B’s PC.
These two results of the first mover with price-setting power only case contrast
with the case of the TIOLI power. There is an important lesson here: when
one of the two parties has price-setting power, but not TIOLI power, she may
use that advantage to advance her distributional interests in a way that implements an inefficient outcome. For example, Ayanda could have implemented
a Pareto-efficient outcome, like point w shown in the figure. This point is on
the Pareto-efficient curve (not shown in the figure), and had she offered Biko
the lower price given by the slope of a budget constraint from point z to point
w, he would have purchased just the amount of coffee that would have implemented point w. But her utility is higher at point n which she can implement by
charging a higher price (steeper budget constraint for Biko).
This explains why it is the case that When Ayanda has price-setting power
only she uses it to get a larger piece of a smaller pie. When she had TIOLI
power she knew that she would get the maximum economic rent (because the
only constraint she faced was Biko’s participation constraint). Subject only to
the participation constraint she could dictate the entire outcome, so she had
no reason to adopt any allocation that was not Pareto efficient.
We will see that this feature of the price-setting case also reappears in other
economic interactions – including credit markets, labor markets, and markets
for goods with limited competition.
Checkpoint 4.7: PSP vs. TIOLI
a. Using Figure 4.13, by reading the relevant points on the x and y axis, say
what the post-exchange allocations for Ayanda and Biko (how much coffee for each, how much data for each). Compare this to the post-exchange
allocations when Ayanda has TIOLI power, calculated in M-Note 4.4.
b. Test your understanding of the first-mover case by explaining the outcome
when Biko is the first-mover. Draw a new version of Figure 4.13.
211
212
MICROECONOMICS
4.12
- DRAFT
Application. Other-regarding preferences: Allocations among
friends
Ayanda and Biko are about to experience one final change in their identities,
along with a personality transplant: they have become friends and they care
about each other. Both are altruistic: they place some positive weight on the
well being of the other. This means, as you will recall from Chapter 2 that they
are other regarding, when evaluating an allocation they take account not only
of the utility they will experience from their bundle but also the utility the other
will experience from their bundle.
They still have a decision to make: how to divide up their coffee (still 10 kilos
of it) and the data (15 gigabytes of it as before). But we will assume now that
neither of them own any portion of either good – so there is no endowment
allocation like our interpretation of point z so far.
The see how the Edgeworth box helps us to understand their decision problem and because this involves some unusual indifference curves, we first treat
a hypothetical case in which Ayanda is completely altruistic and Biko is as
before entirely self-regarding. (We do not imagine that Ayanda would put up
with this, it is just a first step along the way to seeing how two other regarding
friends would look at the problem).
An altruistic utility function
Altruistic Ayanda cares not only about her bundle at an allocation, but also
what Biko gets. Ayanda’s utility therefore depends not only on xA and yA but
also on xB and yB . We measure how much she cares about what Biko gets –
her degree of altruism – by l ( "lambda") a number that varies from 0, if she is
entirely self regarding, to one-half if she places as much weight on what Biko
gets as what she herself gets, in which case she would be called a perfect
altruist.
M-Note 4.7: An altruistic utility function
Remember if Biko did not exist so that Ayanda were making her choice of an allocation in
isolation, her utility would be
uA (xA , yA )
=
(1 a )
xAa yA
But interacting with Biko and dividing goods with him, for l
function as an altruist:
uA (xA , yA , xB , yB )
=
⇣
(1
xAa yA
(4.15)
> 0 we have Ayanda’s utility
⌘
⇣
⌘
a ) (1 l )
(1 a ) l
xBa yB
(4.16)
To see why we say that l is a measure of how much Ayanda cares about what Biko gets
we can take the natural logarithm of equation 4.16
P R O P E RT Y, P OW E R ,
ln(uA )
=
(1
⇣
(1
l )ln xAa yA
a)
⌘
⇣
(1
+ l ln xBa yB
a)
⌘
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
(4.17)
Equation 4.17 says that the natural logarithm of A’s utility is (1
l ) times the natural
logarithm of her valuation (if made in isolation) of her own bundle plus l times the natural
logarithm of B’s evaluation (if made in isolation) of his bundle.
Biko’s utility function a has the same structure as Ayanda’s but the interpretation of l is
the opposite. In Biko’s utility function l is the exponent of Ayanda’s bundle , and (1
l)
is the exponent on his own bundle, the opposite of where these terms appear in Ayanda’s
utility function. The totally self-regarding person, Biko in this case, places no weight on
the bundle of the other person; his degree of altruism, l = 0. So self-regarding B’s utility
function is:
uB (xA , yA , xB , yB )
=
=
⇣
(1
xAa yA
⌘ ⇣
⌘
a) 0
(1 a ) 1
xBa yB
(1 a )
xBa yB
(4.18)
which is just his previous utility function before we introduced l . The rearrangement of
the equation in the second line is true because any term raised to a zero exponent (as in
Biko’s utility function) has a value of 1.
Checkpoint 4.8: Spite and love
.
a. What would it mean in the utility function 4.16 if we had l < 0? Can you give
an example of someone acting as if they had preferences like this?
b. Can you imagine a person having a value of l greater than one half, what
would this mean? Can you think of situations in which people have acted on
preferences of this type?
An altruistic indifference map
To draw her indifference map, we will give Ayanda some particular value of l .
Figure 4.14 shows an Edgeworth box representing a-not-perfectly-altruistic
Ayanda with l = 0.4.
Ayanda’s indifference curves look like the contours on a topographic map of
a mountain. We described the constrained optimization process in Chapter
3 as a kind of hill climbing. where both elements in the bundle were a "good"
and over the entire map, the mountain rose to higher levels if you moved in the
"north east" direction, that is more of both goods. In those figures you never
saw the top of the mountain, because there was not any top. There was no
such thing as "too much" of either good.
But Ayanda’s indifference map has a definite peak at the allocation indicated
by point v. The reason is that from her other-regarding perspective she can
have "too much" of a good when that means that Biko (who she cares about)
too little. This is why Ayanda’s indifference curves oval shaped, just like the
description of a mountain and its peak on a contour map.
213
MICROECONOMICS
9
B's coffee (kilograms), xB
7
6
5
4
3
2
Biko
1
0
k
A's highest utility
v uA
4
uA3
uA2 uA1
j
z
0
Ayanda
1
2
3
4
5
6
7
A's coffee (kilograms), xA
8
9
10
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
10
9
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
8
B's coffee (kilograms), xB
7
6
5
4
3
2
0
k
A's highest utility
uB1
Biko
1
uB2
uB3
v uA
4
uB4
uA3
uA2 uA1
i
Pareto−efficient
Curve
j
z
uB5
0
Ayanda
1
(a) Altruistic Ayanda’s indifference curves
2
3
4
5
6
7
A's coffee (kilograms), xA
8
9
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
B's data (gigabytes), yB
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
8
B's data (gigabytes), yB
A's data (gigabytes), yA
10
- DRAFT
A's data (gigabytes), yA
214
10
(b) Altruistic Ayanda and self-regarding Biko
Notice that when she has little of either good (close to her origin in the lower
left of the box) her indifference curves look as you have seen before. In this
situation both coffee and gigabytes are "goods" so more of each is better, and
the indifference curves slope downward, as you would expect. Moving up or
to the right brings you to a higher indifference curve. In this part of the figure
"more is up."
But beyond a certain point "more" for Ayanda is no longer "up". If she has
most of both goods then getting even more is not something she values,
so moving up and to the right leads her to lower not higher indifference
curves.
To understand the upward-sloping parts of Ayanda’s indifference curves,
remember that if one of the axes represents a good and the other a bad,
then the indifference curve slopes upwards, as in the case of study time (a
bad) and expected grades (a good). In the upper right of the box for example
near point k where she has most of both goods and Biko has little of either
the indifference curves slope downward because for Ayanda having more of
either good (and Biko having less) reduces her utility: both her coffee and her
gigabytes are "bads" not goods.
In panel b of Figure ?? we add Biko’s conventional (self-regarding) indifference curve, so we now know how both of them evaluate every feasible
allocation given by the dimensions of the box. To do this we use Biko’s selfregarding utility function with the value he places on Ayanda’s utility being
zero that is l = 0 because he is entirely self-regarding (that is, zero altruism).
The Pareto-efficient curve is, as before, made of points of tangency between
Ayanda’s and Biko’s indifference curves. But now we exclude tangencies at
allocations for which Ayanda places a negative value on having more of one or
Figure 4.14: Allocation and distribution with
one altruistic person and one self-regarding
person. In panel a the green oval shaped curves
labeled uA are the indifference curves based on
Ayanda’s utility function. In both panels, points z
and i are the same allocations here as in Figure
4.6. Notice that in panel a because Ayanda values
what Biko gets she regards the j as equivalent
to the endowment k, despite the fact that she
receives less of both goods at j than she does at
k. For the same reason, Ayanda’s utility reaches a
maximum at the allocation v indicated in the figure.
The Pareto-efficient curve now does not include k,
because Biko is so deprived of both goods at that
point that Ayanda prefers v to k.
P R O P E RT Y, P OW E R ,
8
7
6
5
4
3
2
Biko
1
0
uB1
uB2
uB3
A
v
uB4
uA3
uA4
uA2
uA1
i
vB
Pareto−efficient
curve
z
0
Ayanda
1
2
3
4
5
6
7
A's coffee (kilograms), xA
8
9
10
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
10
9
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
B's coffee (kilograms), xB
8
7
6
5
4
3
2
1
Biko
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
uB1
uB2
vA
uB3
uA5
Pareto−efficient
curve
uB4
uA4
i
uA3
uB5
vB
uA2
uA1
0
Ayanda
1
2
(a) More altruism (l = 0.4)
both of the goods, above and to the right of her "utility peak" at v. As a result
the Pareto-efficient curve in Figure 4.14 looks different from the one in Figure
4.6 as it does not extend upwards and to the right beyond Ayanda’s maximum
v. Ayanda does not want more of either good than she gets at her maximum
v, while Biko prefers j to any allocation in which she gets less of either or both
of the goods.
Checkpoint 4.9: Altruistic comparisons
Consider Figure 4.14
a. Where is Biko’s utility peak in the figure (analogous to Ayanda’s allocation at
point v?
b. where would point v be if l = 12 (or as close to l = 12 as possible?
c. What happens if Ayanda is self-regarding and Biko is an altruist? How would
the Edgeworth box change?
Efficiency and fairness among altruists
With these analytical tools we can now look at the decision problem faced
by the friends Ayanda and Biko both with other-regarding social preferences.
Figure 4.15 hows for the same Edgeworth box, the indifference maps of the
two.
Unlike the case of one altruistic actor, now both participants have preferred
allocations in the interior of the Edgeworth box. They both would like to avoid
"too much of a good thing."
Both of them dislike extreme allocations giving most to one or the other.
This would not be the case were they evaluating bundles in isolation, that
is if the other person did not exist. The reason why they place a negative
3
4
5
6
7
A's coffee (kilograms), xA
8
215
z
9
B's data (gigabytes), yB
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
B's coffee (kilograms), xB
A's data (gigabytes), yA
9
B's data (gigabytes), yB
A's data (gigabytes), yA
10
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
10
(b) Less altruism (l = 0.2)
Figure 4.15: Altruistic indifference maps. The
two panels depict two different levels of altruism:
high (l = .4) in panel a and low (l = .1) in panel
b. The allocations indicated by the points vA and
vB are respectively A’s and B’s preferred allocation.
The Pareto-efficient curve is composed of all
allocations at which both own coffee and own data
are "goods" rather than "bads" to both A and B, and
where their marginal rates of substitution are equal,
that is, their indifference curves are tangent.
216
MICROECONOMICS
- DRAFT
value on getting more when they already have a lot is not due to diminishing
marginal utility, it is because getting more for yourself means getting less for
the other.
Each of their preferred allocations are shown in the figures by the allocations,
vA for Ayanda and vB for Biko. Around each person’s preferred allocation,
their iso-social welfare curves move outwards and downwards in all directions,
corresponding to lower and lower levels of utility.
As you can see from panel a of Figure 4.15 the Pareo efficiency curve is a line
between their two preferred "utility peaks" vA and vB . By comparing panels
a and b depicting greater and lesser degrees of altruism, you can see that
the more altruistic they are, the shorter the Pareto-efficient curve is, because
greater altruism eliminates more of the extremely unequal allocations.
There is still a conflict of interest, however. At Ayanda’s preferred allocation
Biko has a level of utility less that than the utility he enjoys at this own preferred allocation. The same is true of Ayanda: she does much better at her
preferred allocation than at Biko’s
Along the Pareto efficient curve movements in one direction or the other
necessarily involve one gaining and the other losing. As always the Pareto
efficient curve is a conflict region even among altruists. The fact that the ’utility
peaks’ are closer together in panel a illustrating a greater degree of altruism
means that the conflict of interest between them is lesser the more altruistic
they are.
How might they resolve their remaining conflicts of interest? Here, to make
a decision, they need to go beyond their own utilities (even taking account of
their altruistic nature) to bring in some additional way of making a judgement.
They might adopt:
• a social norm that they both share, for example if one of the two found
the coffee and the data they could go by "finders keepers"; in this case
whichever of them who found the goods could make the decision, presumably implementing his or her preferred allocation.
• a procedural rule of justice, for example flipping a coin to see whose preferred allocation vA or vB would be implemented; or
• a substantive rule of justice, for implementing the allocation recommended
by the Impartial Spectator.
Point i in the figures is a reference point showing the allocation that the Impartial Spectator (who weights Ayanda’s and Biko’s utilities equally) would
implement. This is the same allocation that they would have both preferred
had they been perfect altruists.
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
217
Checkpoint 4.10: Altruism and Rents
Why does altruism reduce the conflict over which allocation to implement?
4.13
The rules of the game and the problem of limited information
Case
Constraints implied
by the rules of the game
Objectives of the actor(s)
Characteristics of the
resulting allocation
Impartial spectator
The available goods
(dimensions of the
Edgeworth box)
Social welfare equally
weighting the utility
of each
Pareto efficient and fair
(by the standard of the
social welfare function).
Symmetrical bargaining
with no first-mover
advantage
Endowment allocation
(private property).
Each player’s participation constraint (PC) at
each stage of
the bargaining
Utility of the two traders
Pareto-efficient if no
impediments to bargaining,
otherwise possible Pareto
improvements over the
endowment allocation
Take-it-or-leave-it power
Endowment allocation
(private property)
Second mover’s
PC
Utility of the first
and second movers
Pareto efficient, first
mover’s rent is
all of the gains from exchange
Price-setting Power
Endowment allocation
(private property)
Second mover’s
incentive compatibility
constraint (ICC)
Utility of the first
and second movers
Pareto inefficient;
first mover gets
most of the gains from exchange
but 2nd mover
gets some
Legislation
The available goods
(money and time)
Whatever the legislators
were seeking to (possibly
the social welfare optimum)
Pareto inefficient,
could be improved
upon by private
bargaining
Bargaining away from
legislated hours
and wages
The new participation
constraints given the
fallback position
implemented by the
legislation
Utilities of the two players
Pareto-efficient if no
impediments to
bargaining, otherwise
possible Pareto
improvements over
the new fallback
Altruism
The available goods
(dimensions of the
Edgeworth box)
Utilities of both (taking
account of how much
they value the other’s
bundle); fairness
Pareto efficient and
(if they can agree
on a fairness
principle) fair.
We have examined several institutional approaches to resolving the conflict
between Ayanda and Biko over allocations of available goods. They all illustrate the dilemma posed in social interactions between:
• The goal of reaching an allocation that is Pareto-superior to the endowment
and possibly even Pareto-efficient.
• The goal of resolving the conflict over the distribution of the resulting eco-
Table 4.1: The rules of the game: Objectives,
constraints and the characteristics of the
resulting allocations.
218
MICROECONOMICS
- DRAFT
nomic rents in a way that is fair.
Table 4.1 summarizes some of the key aspects of the cases we have discussed. Which of the scenarios in the table are relevant in any particular case
depends on the rules of the game for the society of which the players are a
part. How well the rules work depends in important part on whether the actors
have the information that we have attributed to them.
• The Impartial Spectator needs to know a lot about Biko and Ayanda to
implement his preferred outcome, in particular their preferences.
• The symmetrical traders need little information other than their own preferences; they simply continue accepting exchanges as long as the post
exchange bundle is preferable to the pre-exchange bundle.
• The person with TIOLI power needs to know the second-mover’s participation constraint (a single indifference curve), which is less information than
the Impartial Spectator requires.
• The person with price-setting power needs to to estimate the secondmover’s best-response incentive compatibility constraint, which requires
more information than the participation constraint, but less information than
the Impartial Spectator. If the legislator (who imposed the hours and wages
law) was intending to implement an efficient and fair outcome such as the
one recommended by the Impartial Spectator, then he (or they) would have
to know as much as was required of the Impartial Spectator, namely the
entire preference maps of the two.
• The two altruists need to know both their own and the others preferences
(without knowing what the other cares about it is impossible to care about
the other). This is as demanding as the information required of the Impartial Spectator.
A basic fact of economic life is that information is scarce and local. For example in their altruistic friends scenario Ayanda and Biko probably know a lot
about each others preferences, but this is unlikely when Ayanda is the employer and Biko her prospective worker. This will have important ramifications
in the chapters to come especially when we study the labor market, the credit
market, and other exchanges where limited information makes it impossible to
implement Pareto eficient allocations.
4.14
Conclusion
From the silent trade that Ibn Battuta and Herodotus described centuries
ago to eBay, Amazon and Alibaba today, people have exchanged goods to
their mutual advantage and engaged in conflicts over who would get the
lion’s share of the the gains from exchange. The four scenarios we have
H I S TO RY For Friedrich Hayek, an important
20th century economist and philosopher, the
fact that information is scarce and local was
the basis of his criticism of centrally planned
economies – such as the Soviet Union at the
time – and his advocacy of private property
and markets. See his thoughts on this in the
headquote for Part IV of this book and in
Chapter 14.
P R O P E RT Y, P OW E R ,
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
introduced have made it clear that the outcomes of these exchanges and
conflicts depend on the institutions under which they take place, and the
preferences of the people involved. We will see in later chapters that it is
quite common that one of players in an economic interaction has price-setting
power or its equivalents: the power to set wages, interest rates, and other
terms of an exchange.
We have also seen some of the scenarios require that an actor knows a lot
about the other person which is not very realistic even in the two person
case we have used as a simplification of societal interactions. The fact that
information is both scarce and local will play an important role in our analysis
of capitalism as an economic system in later chapters.
The abstract scenarios we have introduced here do not capture the often stepby-step dynamics by which people move from their endowments to eventually
reach an allocation through a series mutually beneficial trades. We will see
that exactly how an economy moves from an out-of-equilibrium endowment
to an equilibrium final allocation makes a difference for the fairness of the
outcomes, but to economists it remains a vexing and far from settled problem.
Making connections
Constrained optimization in strategic interactions: The constrained optimization techniques developed in Chapter 3) are used to better understand
strategic interactions introduced in Chapters 2 and 1.
Optimization rules. In addition to the mrs = mrt rule which we developed in
Chapter 3 for individual optimization we also have the mrsA = mrsB rule
defining a Pareto efficient outcome, both of which are used in strategic
interactions.
Mutual gains from trade: If the endowment allocation (status quo) is not
Pareto-efficient, then mutual gains are possible by implementing some
different allocation of the goods which people may be able to agree to
voluntarily.
Rents and conflicts: These improvements over the fallback option accruing to
the players are rents, made possible by the gains from exchange.
Institutions (rules of the game) and bargaining power: The distribution of
these rents in the Nash equilibrium allocation depend on the players preferences and the initial endowment as well as on the property rights in force,
other institutions and the forms of bargaining power that each participant
can exercise.
Pareto efficiency, institutions: If players have sufficient information some insti-
219
220
- DRAFT
MICROECONOMICS
tutions will result in Pareto-efficient outcomes. Examples are the allocation
implemented by the imaginary Impartial Spectator, and the situation in
which one person has take-it-or-leave-it power. Price-setting power by one
person, however, results in a Pareto-inefficient outcome even with unlimited
information.
Self-regarding and social preferences: Among the set of Pareto-efficient allocations there will generally be conflict of interest among the participants.
But, the extent of these conflicts may be reduced by social preferences
such as altruism.
Important Ideas
utility function
marginal rate of substitution
Cobb-Douglas Utility
Edgeworth box
pareto-criterion
pareto-improving lens
pareto-efficiency
pareto-efficient curve
utility possibilities frontier
endowment
post-exchange allocation
impartial Spectator
social welfare function
mrsA = mrsB rule
iso-welfare curve
altruism
private property
first-mover advantage
take-it-or-leave-it power
price-setting power
participation constraint
incentive compatibility constraint
price-offer curve
institutions
gains from trade
economic rent
Mathematical Notation
Notation
Definition
a
u()
x̄, ȳ
p
W
l
h()
a
exponent of good x in the Cobb-Douglas utility function
utility function
total amounts of x and y available for trade
price of coffee in terms of data
Social Welfare function
extent of altruism
non-linear term of quasi-linear utility function
parameter in the linear term of quasi-linear utility function
Note on super- and subscripts: A, B and i: people; z: endowment point; ti :
outcome with a take-it-or-leave-it offer.
Discussion Questions
See supplementary materials.
P R O P E RT Y, P OW E R ,
Problems
See supplementary materials.
Works cited
& E X C H A N G E : M U T UA L G A I N S & C O N F L I C T S
221
5
Coordination Failures & Institutional Responses
DOING ECONOMICS
This chapter will enable you to do the
following:
Right now, my only incentive is to go out and kill as many fish as I can...any fish
I leave is just going to be picked by the next guy.
2
John Sorlien. Rhode Island (USA) lobsterman
Don’t get him wrong: John Sorlien, the lobsterman, is not the kind of selfinterested and amoral Homo economicus you might find in an economics
textbook. He is actually an environmentalist of sorts, and as President of the
Rhode Island Lobstermen’s Association he was up against a serious problem
of incentives, not a shortcoming of human nature. When he started lobstering
at the age of 22, he set his traps right outside the harbor at Point Judith, within
a few miles of the beach, and made a good living. But the inshore fisheries
have long since been depleted, and now his traps lie 70 miles offshore. He
and his fellow lobstermen are struggling to make ends meet.
Across the world in Port Lincoln on Australia’s south coast, Daryl Spencer,
who dropped out of school when he was 15 and eventually drifted into lobstering, has done much better. During the 1960’s the Australian government
assigned licenses – one per trap – to lobstermen working at the time, and
from that time on, any newcomer seeking to make a living trapping lobsters off
of Port Lincoln had to purchase licenses.
• Understand how the external effects of
our actions on others that are not taken
into account when people make choices
lead to coordination failures.
• Represent social interactions with
graphical and algebraic indifference
curves, feasible sets and best responses
functions.
• Use best response functions to see how
the fairness and Pareto efficiency of the
resulting allocations will depend on the
rules of the game.
• Understand how improvements in
property rights, government policies
such as taxes or direct regulation, the
exercise of power by private individuals
and social preferences can all result in
Nash equilibria that are Pareto-superior
to the Nash equilibrium that would result
in their absence.
• See that the Pareto-improvement made
possible in each of these cases occurs
because (in very different ways) they
induce actors to internalize the external
effects that their actions have on others.
Spencer purchased his start up licenses for a modest sum and by 2000 his
licenses were worth more than a million U.S. dollars (in 2000 prices); considerably more valuable than his boat. More than giving Spencer a valuable
asset, the policy has limited the Australian lobstermen’ s work: Spencer has
60 traps, the maximum allowed; in the Atlantic off of Point Judith John Sorlien
pulls 800 traps and makes a lot less money.
Regulating the amount of lobsters trapped is a coordination problem. Point
Judith and Port Lincoln represent extremes along a continuum of failure and
success; with the lobstermen of Port Lincoln reaping the mutual gains made
Figure 5.1: Sounding the alarm on climate
change, a coordination problem. Greta Thunberg, then 16 years old, speaking at the United
Nations in 2019 about what is probably the most
serious coordination problem that humanity has
ever faced. She said: “We are in the beginning
of a mass extinction, and all you can talk about is
money and fairy tales of eternal economic growth.
How dare you!”1
224
MICROECONOMICS
- DRAFT
possible by a joint decision to limit the number of traps. One may wonder why
the Point Judith fishermen do not simply emulate the Australians. This is especially surprising since one of Sorlien’s friends and a fellow Point Judith lobsterman visited Port Lincoln, returning with tales of millionaire fishermen living in
mansions. But getting the rules right is a lot more difficult than the Port Lincoln
story may suggest, and good rules often do not travel well.
One of the common obstacles to successful coordination is that the rules that
address the coordination problem also implement a division of the gains to
cooperation. In Port Lincoln, those who were awarded the licences benefited;
others did not. Had the young Daryl Spencer not agreed one day to help
out a lobsterman friend and then decided to become a lobsterman himself,
someone else would be a millionaire, and Spencer might still be painting
houses and complaining about the high price of lobsters.
Even if policies to address coordination failures could result in benefits for
everyone affected, how a group coordinates, and what policies they coordinate
on will affect how these benefits will be distributed. And this makes it difficult
to agree on a policy. An example is the Ultimatum Game experiment, in which
conflicts over the size of the Proposers and Respondent’s "slice of the pie"
sometimes result in neither getting any piece of the pie at all.
Conflicts over the distribution of the gains to cooperation have sunk many
otherwise viable agreements to limit the depletion of fishing stocks. To give an
example, a confederation of tribes of North-west U.S. Native American salmon
R E M I N D E R A coordination problem is a
situation in which people could all be better
off (or at least some be better of and none
be worse off) if they jointly decide how to act
– that is, if they coordinate their actions –
than if they act independently.
fishermen seeking to limit their catch decided to allocate shares of a given
maximum catch to each tribe.
In the course of months of debate and bargaining the following principles
of division were proposed, with each proposal more or less transparently
benefiting one or another tribe or class of people:3
• One tribe one share.
• Shares allocated in proportion to a tribe’s number of members.
• Shares to each tribe based on the tribe’s expenditure on lobbying (seeking
to influence) the U.S. Federal Government to adopt policies more favorable
to the tribes, and finally.
• Shares to each tribe in proportion to the relative quantities of fish taken at
the time of the initial treaty with the U.S. Government.
Neither unrestricted competition nor marketable permits to catch specified
amounts (similar to Australia’s lobstering licenses) was proposed. The variety
of proposals and their different effects on the distribution of income among
the tribes suggest how challenging it may be to agree on a rule for sharing the
gains to cooperation.
P ERMIT A permit allows a firm or person
to engage in an activity: it gives them
permission. A permit gives the holder a
property right to a certain amount of a good
or output. For example, a fishing permit
would allow a certain number of fish to be
caught or a carbon emission permit would
allow a certain amount of carbon dioxide to
be emitted during production. When permits
are transferable, firms and people can buy
and sell permits at a price.
C O O R D I N AT I O N F A I L U R E S
Depleting a fishing stock is little different in the structure of its incentives
and its consequences from many other social interactions. In Chapter 9,
for example, using exactly the model we develop here of the coordination
problem that fishers face "harvesting fish," we will study how firms compete
on markets "harvesting customers" by attempting to charge lower prices than
their competitors.
What do firms competing on markets have in common with fishing people
depleting the basis of their livelihood? The common idea is over-harvesting
– whether it is fish or customers – that could be prevented if the firms or
fishermen coordinated their actions rather than acting singly. Just as the Port
Lincoln lobstermen discovered that they could benefit by making a common
decision to limit the number of traps they set, so too will firms discover that
they could make higher profits if they were able to agree on a price at which to
sell, rather than competing.
In Chapter 9 we will return to the fact that coordination among the firms to set
a common price – which is illegal in many countries – raises profits but harms
buyers and as a result increases inequality.
The fact that coordination problems take so many familiar forms explains
both the continuing interest in Hardin’s “Tragedy of the Commons” introduced
in Chapter 1 as well as the impressive amount of human ingenuity that has
been invested in finding ways to avoid or mitigate the costly consequences of
uncoordinated individual optimization in these situations.
In this chapter we develop tools to understand the nature of coordination
problems like the Tragedy of the Commons. We use these tools to analyse
some of the policies (changes in the rules of the game) that improve the Nash
equilibrium outcome when external effects are present.
5.1 Common property resources, public goods, and club goods
Coordination problems are common because when we interact with others
we affect their well being – positively or negatively – and these external effects are not taken into account when we decide on a course of action. The
nature of these external effects differs depending on the type of interaction in
question. In the case of over fishing or ‘over-harvesting’ consumers, when one
person fishes more, or a firm cuts prices, the external effects – on the catch of
the other fishermen or the profits of other firms – are negative.
A taxonomy of goods
To better understand the kinds of coordination problem that we face and how
we might design effective remedies, we classify goods according to their the
& INSTITUTIONAL RESPONSES
225
226
MICROECONOMICS
- DRAFT
kinds of external effects associated with them and the reason why these are a
problem. To do this we ask two questions, introducing two new terms:
• Is the good rival or non-rival?
• Is the good excludable or non-excludable?
When a good is rival, the benefits of its use are limited: more people using the
good reduces the benefit available to others. Your phone is a rival good (our
using it precludes others using it at the same time) while information typically
is non-rival (the fact that I know what time it is and share this information with
you does not preclude your benefiting from the same information).
The distinction between rival and non rival goods can be dramatized by considering how different the reaction would be if you met someone in the street
and politely asked:
• "Excuse me could you give me the time of day?" or
• "Excuse me, could you give me your phone?
When a good is excludable a potential user may be denied access to the good
(or excluded from its usage) at low cost. Your home is an excludable good.
The music from an outdoor concert in a park is not excludable.
We make use of these distinctions to provide the taxonomy shown in Table
5.1. The four categories shown there are "pure cases" introduced to clarify
distinctions. In reality many goods or resources have some aspects of a public
good (they may be a little bit rival and a little bit excludable). The same is true
of the other three categories.
Non-excludability and external effects
But if we just think about the pure cases for now, we have the following:
Common property resources are rival and non-excludable, like in the Fishermen’s Dilemma in Chapter 1. As was the case for the lobstermen above,
the more one fished, the less others caught; but in the absence of a permit
system like they adopted in Australia, no fishermen could be stopped from
fishing, so the common property or pool (the lake or the ocean) was nonexcludable.
Examples of common property resources and their associated coordination
problems include congestion in transportation and communications networks,
overuse of open access forests, fisheries, water resources. Status is another
common property resource, not everyone can be high status (there is a limited
amount to go around) so it is rival. But nobody can be excluded from acquiring status symbols and engaging in other social climbing activities.
C O O R D I N AT I O N F A I L U R E S
Rival
Non-rival
& INSTITUTIONAL RESPONSES
Excludable
Non-excludable
Private good
Common Property (Pool) Resource
(clothing, food)
(fishing stocks, potential buyers),
Club Good
Public Good
(streaming music, online movies)
(global climate, rules of calculus)
A public good is both non-rival and non-excludable. A private good is neither:
it is both rival and excludable.
A slice of pizza is a private good: it is rival because if you eat it nobody else
can enjoy it. It is excludable because the pizza seller can exclude you from
eating it if you do not pay for it. By contrast, weather forecasts (on your phone,
website, or the radio) are a public good. As more people use the weather
forecasts the benefits that those already using the forecasts receive do not
decrease, the benefits of the weather forecasts are non-rival. No person can
be excluded from access to the information about the weather, therefore the
benefits are non-excludable.
When a person contributes to a public good – for example by producing some
new information of value to everyone – she is contributing benefits to others,
so she confers external benefits on others. The problem here is that the
person does not benefit from the positive external effects that her actions
convey on others. So unless the actor values the well being of others as
much as hers own (very unlikely) the public good will be under-provided.
Common pool resources, as the lobsterman John Sorlien explained, will be
over-exploited.
In contrast with public goods and common property resources, there are
"club goods." Club goods are non-rival, but people can be excluded from
their consumption. Common examples include collecting a toll on a little used
highway, charging admission to an uncrowded museum, or making people pay
for streaming video and music.
Intellectual property rights such as patents and copyrights are club goods:
allow people to be excluded from the use of information, which in the absence
of the intellectual property rights would be a public good. This makes it clear
that how some good or resource is classified in our two-by-two taxonomy
depends not only the nature of the good itself, but also on the rules of the
game that determine whether it is excludable or not.
In this chapter we illustrate how coordination failures occur and how policies
might address them with the example of common property resource problems
(or common pool resource problems). The "common property" or "common
pool" is the stock of fish available for catching or the pool of customers who
might purchase the goods sold by the firms. Because common property
resources are non-excludable and rival, people who use them impose external
227
Table 5.1: Public, private, common property and
club goods. In parentheses are examples of the
kinds of goods.
228
MICROECONOMICS
- DRAFT
Figure 5.2: he Grand Banks (North Atlantic)
fisheries: cod landings in tons (1851-2014).
In the 1960s new fishing technologies allowed a
dramatic increase in cod fish caught ("landings")
far outpacing the capacity of the fish to reproduce.
This led to a partial collapse of the fishery in the
1970s and a total collapse in 1992 when the
Canadian government banned fishing entirely.
The fishing stocks to sustainable levels of the past
by the 2030s. Source: Ecosystems and Human
Well-being: Synthesis. A Report of the Millennium
Ecosystem Assessment (2005)4
Landings (tons of cod)
600000
400000
200000
0
1850
1870
1890
1910
1930
1950
1970
1990
2010
Year
costs on each other. The "problem" is that self-regarding people will overexploit the resource because they will not place any value on the negative
external effects of their actions on others. Just such a pattern of exploitation
is shown in Figure 5.2, which displays the catches of cod fish in the North
Atlantic fisheries.
Checkpoint 5.1: A taxonomy of types of goods
Look again at Table 5.1 think of at least two further examples for each of the four
categories of goods.
5.2 A common property resources problem: Preferences
Let’s consider a specific example of a common property resource problem:
the over-exploitation of an environmental resource. It could be the oceans,
or forests, or a livable planet, but we’ll stick to the problem of over-harvesting
fish.
Our questions look at the ways that the rules of the game and the preferences of the actors determine what we should expect to happen in these
situations.
Preferences over fishing time and fish consumed
We turn now to the the problem confronted by two fishermen, called Abdul
(A) and Bridget (B). We model just two fishermen as a way of representing
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
229
Figure 5.3: Abdul and Bridget trying to catch the
same fish. The lake is a common pool resource, so
the benefits are rival and each imposes a negative
external effect on the other.
how a large number of them might interact. They fish in the same lake, using
their labor and their nets. To start, we assume they consume the fish they
catch (what we call their "catch") and do not engage in any kind of exchange.
We will begin by assuming that they do not make any agreements about
how to pursue their economic activities. (Recall that this means that they are
engaged in a non-cooperative game.)
Each derives well-being from eating fish and experiences a loss of well-being
(disutility) with additional fishing time. We represent their preferences when
they are engaged in some amount of fishing with the following quasi-linear
utility functions:
Fisherman’s utility
Abdul’s utility
Bridget’s utility
= Fish consumption
1 A 2
uA (hA , yA ) = yA
(h )
2
1 B 2
uB (hB , yB ) = yB
(h )
2
Disutility of fishing
(5.1)
(5.2)
The utility function given by Equation 5.1 tells us three things about Abdul’s
preferences:
• Consumption (yA ) measured in pounds of fish is a “good;" Abdul derives
utility from obtaining more consumption (consuming more fish) which is
why yA has a positive sign.
• Time spent fishing (hA ) measured in hours is a “bad": the second term has
a negative sign.
• Utility (uA ) is increased by one unit if he is able to consume one more
pound of fish, so the units in which we can measure utility are pounds of
fish.
• Marginal utility is not diminishing but instead is a constant (equal to 1,
D ISUTILITY OF WORK Working doesn’t
only take up time, it is also costly to people
because of the effort that they need to exert.
Manual labor is physically tiring and often,
with activities like construction and mining,
can be dangerous as well as complex and
challenging mentally. Working as a waitress
burns as many calories in an hour as doing
construction work. Office work, too, requires
effort, requiring concentration and attention.
Exerting this effort often isn’t pleasant and
therefore results in disutility or a cost of utility
to exert.
230
MICROECONOMICS
- DRAFT
uA3
uA2
Consumption, yA
slope = hA = 15
Figure 5.4: Abdul’s indifference curves over
output (yA ) and fishing time measured in hours
(hA ). Output (fish) (yA ) is a “good" and provides
Abdul with positive utility, whereas fishing time (hA )
is a “bad". Notice that Abdul’s indifference curves
in fishing hours and output are upward-sloping,
similar to the indifference curves over money
(income, a good) and working time (a bad) in
Chapter 4.
uA1
slope = hA = 10
uAz
g
yg = 247.5
f
yf = 190
z = 112
0
5
hAf = 10
hAg = 15
20
24
A's hours, hA
because an additional pound of fish provides him with a one unit increase
in utility).
Bridget’s utility function Equation 5.2 is interpreted in the same way as Abdul’s. Both of them refer to some given time period, such as a week. So output
and consumption are pounds of fish caught and eaten in a week, while time
spent fishing is hours fished over the course of a week.
If for some reason they do not fish at all, they receive a small amount of fish yz
from neighbors or the government, labeled with the subscript z because this is
their fallback position (as the endowment allocation was in Chapter 10)
To decide how much time to fish, people like Abdul have to balance their
disutility of hours of work with the utility of consumption that they get from
consuming the fruits – or the fish – of their work time. To understand this
process, we look at Abdul’s indifference curves.
Four indifference curves derived from Abdul’s utility function, equation 5.1, are
presented in Figure 5.4. Notice that:
• the higher numbered (meaning more preferred) indifference curves are
above (more fish) and to the left (less work);
• the curves slope upwards because fish is a good and fishing time is a
bad, so comparing points f and g he is indifferent between fishing less and
consuming less (point f) and fishing more and consuming more (point g)
R E M I N D E R The indifference map provides
information on how he evaluates all of the
imaginable combinations of fishing time and
fish caught. It says nothing about the actions
and outcomes that are feasible for him.
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
231
• the lowest indifference curve is labeled uA
z and its vertical axis intercept
is point z or the level of utility measured in fish per week, yA
z that he will
receive if he does not fish at all; and finally
• for any given level of yA the indifference curve is steeper the more hours
Abdul works: the more he works the greater is his dislike of working more
compared to how much he likes eating more fish.
The negative of the slope of his indifference curve is Abdul’s marginal rate of
substitution between fish (yA ) and fishing time (hA ), the ratio of his marginal
utility of fishing time to his marginal utility of fish. This quantity takes a particularly simple form in this case because (as is shown in M-Note 5.1). Abdul’s
marginal utility of fish is 1 and his marginal utility of fishing time is
hA . So,
the marginal rate of substitution of fish consumption for fishing time is:
mrsA (hA , yA )
hA
=
(5.3)
Abdul’s marginal rate of substitution of fish consumption for fishing time is
hA , and this is also his marginal disutility of fishing time, which is negative,
because he regards fishing time as a “bad.”
If he were already working 12 hours, then his disutility of hours of fishing
(which is hA itself) is the greatest amount of fish he would be willing to give
up in order to be able to work an hour less. This is his willingness to pay (in
fish) to have more free time. Or if Abdul were employed where he is already
working hA and paid a wage, then the quantity hA is the lowest wage (paid in
fish) in return for agreeing to work an extra hour, that he would accept.
M-Note 5.1: The mrs(h, y) with quasi-linear preferences
When Abdul’s utility is given by Equation 5.1, we have.
Marginal utility of consuming fish
=
∂ uA (hA , yA )
=1
∂ yA
(5.4)
Marginal disutility of fishing time
=
∂ uA (hA , yA )
= hA
∂ hA
(5.5)
The marginal utility of fishing time is negative (it reduces Abdul’s utility and is equal to
hA ) and we use the term marginal disutility of fishing time for the same quantity but with
a positive sign (it increases Abdul’s disutility).
The marginal rate of substitution of output for hours of work (mrsA (hA , yA )) is the negative
of the slope of the indifference curve, which is ratio of the marginal utilities:
mrsA (hA , yA )
=
hA
=
1
hA
(5.6)
So the slope of the indifference curve is the marginal disutility of working time, or just hA
the amount of working time itself.
A
A
Furthermore, along any of the indifference curves, uA
1 , u2 and u3 , the vertical intercept is
the amount of utility in fish if they were not working at all, that would be the same as the
utility at every point on that indifference curve.
The marginal rate of substitution of Abudl’s fish consumption for Abdul fishing time
M - C H E C K Abdul’s utility function in fish and
fishing time is quasi-linear : since it is linear
in fish – he derives a positive and constant
marginal utility from consuming fish – but is
negative (it is a disutility) and non-linear in
fishing time. His marginal disutility of fishing
time it is not constant, it is greater the more
time he spends fishing.
232
MICROECONOMICS
- DRAFT
mrs(hA , yA ) is an entirely different quantity than the marginal rate of substitution given
by indifference curves for the fishing times of the fishermen mrs(yA , yB ) that we introduce
later.
Checkpoint 5.2: The lake as a common property resource
a. Explain why the lake that Abdul and Bridget are fishing is a common property resource. What are its characteristics? Explain.
b. Return to Chapter 1 and the choice of strategies that the fishermen had
in the Fishermen’s Dilemma to Fish 10 hours or Fish 12 hours. Substitute
these values into the utility functions to see what the payoffs in the corresponding game table would be if the fishermen could only choose these two
strategies. Find the Nash equilibrium of the game.
5.3 Technology and environmental limits: The source of a coordination failure
A coordination problem arises because Abdul or Bridget fishing more reduces the amount of fish the other catches in an hour of fishing. This negative
T ECHNOLOGY A technology is a description
of the relationship between inputs –including
work, machinery, and raw materials – and
outputs.
external effect that each has on the catch of the other is the source of the
coordination problem.
These external effects are part of the technology of fishing. A technology
is a description of the relationship between inputs – such as fishing time,
equipment, and fish in the wild – and outputs – in this case caught fish. A
technology is often described mathematically in a production function. You already used a production function in Figure 3.9 where the input was time spent
studying and the output as learning. (We postpone a detailed discussion of
production functions until the next chapter).
??
Here are the production functions for Abdul and Bridget, where xA for Abdul
and xB for Bridget represents the number of fish caught by each of them
in a week and hA and hB are the hours of fishing time they work during the
week. The production functions translate the actions taken by the two – their
fishing hours (hA and hB ) – into the amount that each catches (xA and xB ) and
consumes (yA and yB ).
Abdul’s catch:
xA (hA , hB ) = hA (a
b (hA + hB ))
(5.7)
Bridget’s catch:
xB (hA , hB ) = hB (a
b (hA + hB ))
(5.8)
The two parameters of the production function are:
• a (Greek alpha) is the the fisherman’s maximum average productivity,
that is, total catch divided by time spent fishing which would occur if one of
P RODUCTION F UNCTION A technology is
a way of transforming inputs into outputs,
described mathematically as a production
function.
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
233
them fished some small amount of time and the other did not fish at all. We
assume that a > 0 otherwise they could not ever catch any fish.
• b (Greek beta) measures the decrease in average productivity for each
hour fished in total by the two. We assume that b > 0 to reflect the fact
their interdependence and the negative external effect that each fishing has
on the other’s catch.
Were we to be applying this model to real fishermen, we would find that b
and a differ. For example, one may have a larger boat and for this reason
may catch more fish in an hour and also have a larger effect on the fishing
productivity of the other. However, because our interest here is not in the
effects of differing sizes of their boats, but instead on differing amount of time
fished, we assume that b and a are the same for the two fishermen.
From Equation 5.7 you can see that Abdul’s total catch is his hours of fishing
multiplied by his total catch per hour of fishing, termed the average productivity of his fishing time. Dividing both sides of Equation 5.7 by hA we have his
average productivity of fishing time:
xA
hA
= a
b (hA + hB )
(5.9)
M - C H E C K We adopt parameters for the
production functions so that Bridget and
Abdul cannot work so many hours that their
average productivity becomes negative, so
that fishing more would reduce their total
catch. This is why we do not extend the
lower of the two production function curves
in Figure 5.5 beyond 24 hours, the point after
which the function turns downwards.
What this means is that:
Average productivity = Maximum
Reductions due to own and other’s fishing time
We are also interested in what is termed the marginal productivity of Abdul’s
fishing time. This is the effect of fishing a little more on the size of his total
catch. In the M-check we show that Abdul’s marginal productivity of fish time
is:
mphA
= a
b (hA + hB )
b hA
(5.10)
This equation can be read as:
Marginal productivity = Average productivity
Reduction due to own fishing time
We call a the maximum productivity of fishing because it is the amount that
would be produced per unit of fishing time when there is no fishing being
done.
The parameter b expresses three important aspects of the technology:
• Decreasing average productivity: If Abdul spends more time fishing, his
catch will be larger, but his average productivity – the size of the catch per
hour fished – decreases.
• Decreasing marginal product of work time : If Abdul already fishes a lot,
then the additional amount of fish that he catches were he to fish a little
more will be less than if he were initially fishing a lesser amount.
M - C H E C K The marginal product of Abdul’s
fishing time is found by partial differentiating
his total catch (xA ) given by the production
function (Equation 5.7) with respect to his
working time (hA ), which gives us Equation
5.10 We study the mathematical and
conceptual properties of production functions
and marginal products in the next chapter.
MICROECONOMICS
Consumption, yA (pounds, lb)
234
- DRAFT
Figure 5.5: Abdul’s production of fish with hours
of fishing and marginal benefit of hours spent
fishing. In the top panel, Abdul’s light green total
product line corresponds to when Bridget does not
fish (hB = 0) and Abdul’s light green total product
line corresponds to when Bridget fishes 12 hours
(hB = 12). Similarly, in the lower panel, Abdul’s light
green marginal benefit line corresponds to when
Bridget does not fish (hB = 0) and Abdul’s light
green marginal benefit line corresponds to when
Bridget fishes 12 hours (hB = 12).
y(hA, hB = 0)
k
288
y(hA, hB = 12)
j
216
0
3
6
9
12
15
18
21
24
27
30
27
30
A's marginal benefit (pounds, lb)
30
24
k
18
Marginal benefit
when, hB = 12
0
3
6
Marginal benefit
when, hB = 0
j
12
9
12
15
18
21
24
A's hours, hA
• Interdependence: The fact that hA appears in Bridget’s production function
and hB in Abdul’s represents the external effects and therefore the interdependence between the fishermen. The fact that the sign of these terms is
negative means that the external effect is negative.
We depict Abdul’s production function in the top panel of Figure 5.5. The
higher of the two green curves represents the relationship between his labor
input and his fish output when Bridget is not fishing at all, that is: hB = 0. (We
will explain the lower curve in a moment.)
Abdul’s production function is increasing but becomes flatter the more time
Abdul fishes. The slope of this curve is the marginal product of time fishing, indicating for each level of hA the increase in the amount of his catch that would
result if he increased his fishing time a little. We also call this the marginal
benefit of fishing time because it indicates how much he benefits if he fishes
a little more (how much the larger catch from additional fishing time raises his
utility).
The second, lower, dark green curve in the top panel shows how Bridget’s
fishing for 12 hours reduces the amount of fish Abdul will catch for each
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
235
level of Abdul’s fishing time. Like the top curve, it is rising: as Abdul spends
more time fishing he catches more fish. But two things about it are important:
• it is lower than when Bridget does not fish, and
• its slope is also lower (it is flatter for each hour that Abdul spends fishing).
Both are the result of the negative external effect that Bridget’s fishing inflicts
on Abdul.
In the lower panel of Figure 5.5 we show the marginal product of an hour of
fishing based on the production function shown in the top panel, labeled the
marginal benefit of hours of fishing.
In Figure 5.5 when Abdul fishes 12 hours a week (and Bridget does not fish),
his catch is 288 but when she also fishes 12 hours (the lower green curve) his
catch is just 216 lbs. Equally important, when Bridget is not fishing, and Abdul
is fishing 12 hours, his marginal product is 18. The fact that the marginal
benefit curve shifts downward when Bridget fishes 12 hours reflects the fact
that in the top figure for any given amount of fishing time by Abdul, the curve is
flatter.
M-Note 5.2: Numerical Examples for Productivity and External Effects
Throughout the chapter, we’ll cover a worked example where Abdul and Bridget have the
same level of productivity and external effect on each other. We shall assume that a = 30
and that b = 12 .
Abdul and Bridget’s utility functions therefore become the following:
Abdul’s utility:
uA (hA , hB ) = hA (30
Bridget’s utility:
uB (hA , hB ) = hB (30
1 B
(h + hA ))
2
1 A
(h + hB ))
2
1 A 2
(h )
2
1 B 2
(h )
2
(5.11)
(5.12)
In the case where the fishermen fished alone, that is the other fishermen had zero hours
fishing, Abdul’s utility would therefore be: uA = 30hA 12 (hA )2 12 (hA )2 = 30 (hA )2 .
When Abdul and Bridget both spend time fishing, the external effect reduces Abdul’s utility, therefore he would have uA
30hA
1 A B
2h h
= 30hA
1 A B
2h h
1 A 2
2 (h )
1 A 2
2 (h )
=
(hA )2 .
5.4 A best response: Another constrained optimization problem
To understand the Nash equilibrium of the interaction between Abdul and
Bridget we will need to know how each will best respond to any of the possible
levels of fishing chosen by the other. This is because a Nash equilibrium is a
mutual best response. To do this we will derive the best response function of
each. But to do this we begin, as we did in Chapter 4, with a simpler problem:
R E M I N D E R A player’s best-response
function gives, for every possible set of
strategy chosen other players player,
the strategy that maximizes the player’s
utility. A strategy profile in which all players
are playing a best response, is a Nash
equilibrium.
236
MICROECONOMICS
- DRAFT
showing how one of them, Abdul, will choose how many hours to fish, when
Bridget is fishing at some given number of hours.
Abdul, choosing a level of fishing time
As a first step we bring together the information we have from the previous
two sections on their preferences and their technology in a single equation
expressing the benefits and costs of fishing:
Utility = Total benefit (fish caught and consumed)
Total cost (disutility of fishing time)
So, for each of them, we substitute the production functions (Equations 5.7
and 5.8) for yA = xA and yB = xB into their utility functions for their total
consumption (Equations 5.1 and 5.2). Doing this we obtain for each of them a
single function showing how their utility depends on their own and the other’s
fishing times, which is why we write their utility functions as uA (hA , hB ) and
uB (hA , hB )
Abdul’s utility:
uA (hA , hB ) = hA (a
b (hA + hB ))
Bridget’s utility:
uB (hA , hB ) = hB (a
b (hA + hB ))
1 A 2
(h )
2
1 B 2
(h )
2
(5.13)
(5.14)
For concreteness let’s suppose that Bridget is not fishing at all: she is a farmer
and does not interact with Abdul in any way. We can therefore substitute
hB = 0 into Abdul’s utility function, Equation 5.13. Then Abdul’s constrained
optimization problem is to maximize his utility subject to the constraint given
by how productive his fishing time is when Bridget is not fishing.
This problem is set out in Figure 5.6 which combines Abdul’s indifference
curves from Figure 5.4 with his production function (when hB = 0) from
5.5.
Abdul might first consider fishing six hours, with results indicated by points
f, g, and h in Figure 5.6. To determine if he should fish 6 hours he would
compare:
• the marginal cost of working more: namely the marginal disutility of working
time, which is the slope of the indifference curve at f shown as point h in
the lower panel with
• marginal benefit of working more: namely, the marginal productivity of his
fishing time, which is the slope of the production function at f shown as
point g in the lower panel.
From either the two slopes at point f (mrs 6= mrt ) in the top panel or their
representation by points g and h in the bottom panel (mb > mc) Abdul would
see he would increase his utility by working more than 6 hours.
How much more? He will adopt the following method. He will compare:
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
237
Consumption, yA (pounds, lb)
Figure 5.6: Abdul maximizes his utility subject
to the constraint of his production function
when Bridget does not fish.
s
300
uA3
225
uA2
144
Feasible set, hB = 0
f
uA1
0
A's marginal costs & benefits (lb)
A's production
function
xA(hA, hB = 0)
defines feasible
consumption, yA
3
6
9
12
15
18
21
24
27
30
Marginal cost
or disutility
mcA = hA
α
g
24
s
15
mb = mc
Marginal benefit
when, hB = 0
6
h
0
3
6
9
12
15
18
21
24
27
30
A's hours, hA
Marginal benefit (mb): The additional fish he would catch is the marginal
benefit of fishing more.
Marginal cost (mc): The additional effort he exerts or disutility it costs him is
the marginal cost of fishing more.
Abdul will best respond if he follows some simple rules:
mb > mc If the marginal benefit exceeds the marginal cost as at point f, then
fish more.
mb < mc If the marginal cost exceeds the marginal benefit, then fish less.
mb = mc If the marginal cost equals the marginal benefit, do not change how
238
MICROECONOMICS
- DRAFT
much you fish.
5.5 A best-response function: Interdependence recognized
You can confirm from the figure that following the rule in italics just above,
Abdul will fish 15 hours if Bridget is not fishing, indicated by point s in the
figure (s for "solo" because Bridget is not fishing). This gives us just one point
on his best response function hA (hB = 0) = 15 hours.
What about when Bridget is fishing, for example, fishing 12 hours? Abdul’s
reasoning is identical to the above rule. This case is illustrated in Figure 5.7
where the new feasible set constraining Abdul is smaller, because his catch
for any amount of time that he spends fishing is reduced by Bridget also
fishing.
Abdul knows that the level of fishing that will maximize his utility under these
new conditions is that which equates:
• the slopes of an indifference curve and his production function so that the
two are tangent in the top panel
• or, to put it another way, the marginal benefit and the marginal cost of more
fishing in the bottom panel.
This gives us a second point on Abdul’s best response function, hA (hB =
12) = 12. Abdul fishes less when Bridget fishes more. This occurs because
Bridget’s fishing more reduces the marginal benefit to Abdul’s fishing.
What about Abdul’s response to Bridget fishing different hours. We do not
have to go through the above process, tediously making a separate figure for
each level of fishing time she might choose.
Instead we can use mathematical expressions for the marginal costs and
benefits of fishing to determine Abdul’s best response not as a discrete point,
but as a continuous function, giving us his fishing time for any level of fishing
Bridget might do.
Using the rule that the best response is the number of hours that equates
marginal benefits to marginal costs we have a general rule that can be expressed mathematically and which allows us to isolate hA as a function of hB
and the parameters a and b . Here is the rule: a best response is a value of
hA that satisfies the following rule:
Marginal benefit
a
A
B
b (2h + h )
= Marginal costs
= hA
(5.15)
Re-arranging Equation 5.36 to isolate hA and to express his optimal fishing
C O O R D I N AT I O N F A I L U R E S
Consumption, yA (pounds, lb)
xA(hA, hB = 0)
300
uA3
225
uA2
144
uA1
A's marginal costs & benefits (lb)
0
uA3
uA2
s
uA1
xA(hA, hB = 12)
n
Feasible set, hB = 12
3
6
9
12
15
18
21
24
Marginal benefit
when, hB = 0
30
Marginal disutility
mcA = hA
24
Marginal benefit
when, hB = 12
15
s
n
12
0
3
6
9
12
15
18
21
24
A
A's hours, h
hours as a function of Bridget’s hours hA (hB ), we have
Abdul’s best-response function:
hA (hB ) =
a b hB
1 + 2b
(5.16)
How does Abdul’s fishing time hA change when the variable (hB ) and parameters (a and b ) change?
• Change in Bridget’s fishing time (hB ): If Bridget decreases her fishing
time, Abdul’s marginal benefit curve shifts up, and Abdul’s best response
is to increase his fishing time to balance his marginal cost with the higher
marginal benefit. Abdul’s best-response function does not shift, he chooses
a different level of fishing due to the change in Bridget’s fishing time.
& INSTITUTIONAL RESPONSES
239
Figure 5.7: Abdul maximizes his utility subject
to the constraint of the production function
when Bridget spends 12 hours fishing. The
feasible set is now smaller because of the negative
external effect that her fishing imposes on Abdul.
In the top panel, at point n his indifference curve
labeled uA1 is tangent to his production function,
meaning in the lower panel, that the marginal
disutility of fishing time is equal to the marginal
productivity of fishing time, or the marginal cost of
fishing more is equal to the marginal benefit.
240
MICROECONOMICS
- DRAFT
• Change in maximum productivity (a ): If Abdul’s basic productivity increases, and nothing else changes, this shifts his marginal benefit curve up
and independently of any change in Bridget’s fishing time, he will increase
his fishing time to balance his marginal cost with the higher marginal benefit. This is a shift in Abdul’s best-response function itself, not just a movement from one point on it to another as in the bullet above.
• Change in the over-fishing effect (b ): If the external effect increases,
Abdul’s marginal benefit curve pivots downward with a corresponding
decrease in fishing time (b changes the slope of his marginal benefit curve,
as can be seen from Equation 5.36). Like the increase in a , in this case
Abdul changes his fishing time due to a shift in this best-response function.
The best-response function for Bridget can be derived in the same way we
derived Abdul’s. Therefore her best-response function is:
Bridget’s BRF :
hB (hA ) =
a b hA
1 + 2b
(5.17)
M-Note 5.3: Marginal benefits, marginal costs, and finding the best responses
In M-Note 5.2, we used the example of a = 30 and b = 12 to provide utility functions for
Abdul and Bridget, as represented in equations 5.11 and 5.12. We now use those parameters to identify the first-order condition for Abdul’s utility maximization where his marginal
benefits equal his marginal costs and therefore to provide a best-response function.
uA (hA , hB )
=
hA (30
∂ uA
∂ hA
=
(30
|
uAhA =
1 B
(h + hA ))
2
1 B
h
2{z
hA )
}
Marginal benefit
1 A 2
(h )
2
hA
|{z}
=0
Marginal cost
We can isolate Abdul’s hours of work, hA to find his best response to Bridget’s hours of
work:
Abdul’s BRF:
Bridget’s BRF:
hA (hB )
hB (hA )
=
=
30
2
30
2
1 B
2h
1 A
2h
= 15
= 15
1 B
h
4
1 A
h
4
(5.18)
(5.19)
Each of them therefore has a best-response function that is a function of the other person’s time spent fishing: hA (hB ) for Abdul and hB (hA ) for Bridget.
M-Note 5.4: Mathematics of the best-response function
To understand each player’s response to the other, it is useful to understand their marginal
utilities of hours of fishing. We do this for Abdul, in the understanding that the Bridget will
have symmetrical results. We will therefore find uA
, Abdul’s marginal utility of his own
hA
hours of fishing, uA
, marginal utility of Bridget’s hours of fishing, and hA (hB ), Abdul’s
hB
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
241
best-response to Bridget’s choice of hours. We start with Abdul’s utility function:
uA (hA , hB ) = hA (a
b (hA + hB ))
1 A 2
(h )
2
We can differentiate Abdul’s utility function with respect to his own hours (hA ) to find his
marginal utility of his own hours of work. We also differentiate his utility function with respect to Bridget’s hours of work to find how his utility changes when Bridget changes her
hours (hB ).
A’s marginal utility of hA
uAhA =
Marginal effect on A’s utility of hB
uAhB =
A
If we set Abdul’s marginal utility uA
= ∂∂ uhA
hA
Bridget’s hours of work:
∂ uA
∂ hB
=
a
b hB
2b hA
=
a
b hB
hA (1 + 2b )(5.21)
=
=
a
b hB
=
a
b hB
hA
=
a b hB
(1 + 2b )
A s BRF:
hA (5.20)
b hA
(5.22)
= 0, then we can find his best response to
∂ uA
∂ hA
hA (1 + 2b )
uAhA =
Isolate hA term
∂ uA
∂ hA
hA (1 + 2b ) = 0
Which is what we found from setting marginal benefit equal to marginal cost to find Abdul’s
best-response function in Equation 5.16.
We can use the above to define Abdul’s marginal rate of substitution:
mrsA (hA , hB )
=
=
uAhA
uAhB
a
b hB
(1 + 2b )hA
b hA
(5.23)
Checkpoint 5.3: How the BRFs change
Consider how the parameters would change the best response functions in
M-Note 5.2.
a. What would happen if a = 30 changed to a = 24 and b = 12 changed to
b = 13 ?
b. What would the vertical and horizontal intercepts be for each of Abdul and
Bridget? Sketch the best response functions you found in a.
5.6 How will the game be played? A symmetric Nash equilibrium
We do not have enough information to answer the question in the section
title. To do this we need answers to other questions. Is one of them powerful
enough determine the allocation unilaterally, stating: I fish 15 hours, and you
are excluded from fishing? Is there a government that can place a tax on
fishing to discourage over-harvesting the stock? Can Abdul and Bridget agree
to fish less? If they did, can they count on their agreement being enforced? In
other words, we need to know more about the rules of the game.
R E M I N D E R We began our analysis of
Ayanda and Biko trading data and coffee in a
similar way, with the two being symmetrical
traders with neither of them having any
particular advantage in the bargaining
process.
242
MICROECONOMICS
- DRAFT
One possibility is that the two are independent (they do not make agreements
with each other), self-regarding, and symmetrical, in that neither has any
particular advantage in their interaction. So they simply try to do the best that
they can, given what the other is doing and given the information they have.
We will investigate other rules of the game later.
A stationary allocation among symmetric players
To study this case, we graph the two best-response functions in Figure 5.8.
This gives us all the information we need to determine the Nash equilibrium of
their interact on.
A Nash equilibrium is a mutual best response, so Abdul’s choice of fishing
hours must be a best response to Bridget’s choice of fishing hours, which
must in turn be a best response to Abdul’s choice of fishing hours. This
sounds complicated but with a little help from the mathematics we have already done, it is not: A Nash equilibrium is a point that is on both of the players’ best-response functions. We label the point n and define the hours that
they work at the Nash equilibrium as (hAN , hBN ), where point n is the Nash
equilibrium in the figure and the superscript N indicates each player’s Nash
equilibrium hours. A Nash equilibrium is a pair of fishing times (hAN , hBN ) that
satisfy each fisherman’s best-response function.
When both players act according to their best-response functions, the outcome is a Nash equilibrium. In Figure 5.8 we plot the two best-response
functions. The Nash equilibrium is the point where the two best response
functions intersect the only point that the two lines have in common, shown by
hAN and hBN .
We show in the M-Note 5.5 how to find the Nash equilibrium hours of fishing
for each person. At the Nash equilibrium, the two fishermen will spend the
same amount of time fishing.
hAN = hBN =
a
1 + 3b
(5.24)
Equation 5.24 shows that each fisherman’s hours spent fishing is defined by
the parameters a and b , capturing the effects on their best-response of their
average productivity, their decreasing marginal productivity, and the negative
external effect each has on the other.
The Nash equilibrium fishing hours, hAN and hBN , are equal because the
Abdul and Bridget have identical utility functions (other than reversing the superscripts), and they are determined by the parameters a and b . The greater
is the maximum average productivity, a , the greater will be their equilibrium
hours of fishing. The larger is the negative external effect each has on their
own productivity and on the other person’s productivity, b , the lower their
equilibrium hours will be.
C O O R D I N AT I O N F A I L U R E S
B's hours, hB
α
= 15
1 + 2β
h
243
Figure 5.8: Nash equilibrium: mutual best
responses for Bridget and Abdul. The equations
for the best-response functions are:
24
BN
& INSTITUTIONAL RESPONSES
B's best−response
function
●
= 12
n
hAN = 12
0
α
= 15
1 + 2β
24
A's hours, hA
M-Note 5.5: Finding Nash Equilibrium fishing time
By definition of the Nash equilibrium, Abdul’s Nash equilibrium fishing time must be a best
response to Bridget’s Nash equilibrium fishing time, and Bridget’s Nash equilibrium fishing
time must be a best-response to Abdul’s Nash equilibrium fishing time. A Nash equilibrium
is therefore a pair of fishing times (hAN , hBN ) that satisfy the following equations:
hAN = hA (hBN ) =
(a b hBN )
(1 + 2b )
(5.25)
hBN = hB (hAN ) =
(a b hAN )
(1 + 2b )
(5.26)
Equations 5.25 and are two linear equations in two unknowns. We can solve the equations
for the unknowns, which are the fishing times at the Nash equilibrium.
There is a particularly simple way to do this in our case because:
1. The two fishermen have identical utility functions (they are mirror images of each
other); so
2. we know that it must be that hAN = hBN , and
3. we can therefore set the Nash equilibrium level of fishing of the one equal to the
best-response function of the other.
So substituting hB = hA , into Abdul’s best-response function is:
hA (hB )
Multiplying out and isolating hA :
=
a b hA
1 + 2b
a b hA
1 + 2b
hA hB =
a b hB
1 + 2b
If a = 30 and b = 0.5, the parameters we used
in the previous figures, then we can see that when
Bridget does not fish (the intercept of Abdul’s
best response function with the horizontal axis)
he fishes 15 hours. The point at which their bestresponse functions intersect is the Nash equilibrium
of the interaction. Using these same parameters,
we can see that the Nash equilibrium given by
Equation 5.24 is that they both fish 12 hours.
Nash Equilibrium
A's best−response
function
0
hB hA =
244
MICROECONOMICS
- DRAFT
hA + 2b hA
=
a
A
h + 3b h
=
a
hA (1 + 3b )
=
a
AN
=
A
h
b hA
a
= hBN
1 + 3b
5.7 How would the players get to the Nash equilibrium? A dynamic
analysis
When we used the equation for the Nash equilibrium level of hours of fishing
(Equation 5.24) to say what the effect of a change in a or b would be, we
used what is called comparative static analysis.
C OMPARATIVE STATICSWhen using comparative statics we compare the status
quo outcome or the Nash equilibrium before the change with the outcome or Nash
equilibrium after the change.
When using comparative statics we compare the status quo outcome or the
Nash equilibrium before the change with the outcome or Nash equilibrium
after the change.
• The word static refers to the Nash equilibrium because at a Nash equilibrium there are no reasons for the actors to change what they are doing.
• The process is comparative because we compare two or more states
before and after a change.
We did the following:
• We started by assuming that Aram and Bina were at the Nash equilibrium,
each fishing 12 hours.
• We then assumed that other things (like the weather) that might affect their
fishing time are held constant (this is called the ceteris paribus assumption
or ’other things equal’).
• Then we compared the two Nash equilibria, one before the change and the
other after the change, and.
• We assumed that after the change Aram and Bina would be at the new
Nash equilibrium, working some different number of work hours.
• Finally we considered the difference in work hours between the two Nash
equilibria to be the effect of the change on work hours.
This type of analysis gets the name comparative static because it compares
two static (unchanging) situations without looking at how the change takes
place. This is an essential method of economic analysis, simplifying the matter
by a shortcut. The shortcut is that we did not explore the process by which
the move from the first to the second Nash equilibrium occurs, namely who did
what to implement the move. This is called a dynamic analysis because it is
R E M I N D E R When we say "other things
equal" we are using the ceteris paribus
assumption which allows us to compare
what happens when one variable of interest
changes.
C O O R D I N AT I O N F A I L U R E S
Marginal benefit
Marginal disutility
when, hB = 12
mcA = hA
24
mb > mc
A
mbA < mcA
mbB < mcB
mbA > mcA
at 6 hours
12
mbA < mcA
at 18 hours
n
245
A's best−response
function
A
B's hours, hB
A's marginal costs & benefits (lb)
24
& INSTITUTIONAL RESPONSES
mcB < mcB
15
n
12
B's best−response
function
mbB > mcB
mbB > mcB
mbA > mcA
mbA < mcA
0
0
3
6
9
12
15
18
21
24
0
6
A's hours, hA
(a) Marginal benefits and costs when B fishes 12 hours
12
15
18
24
A's hours, hA
(b) Dynamic analysis of A and B’s choices
based on the process of change. (The term dynamic refers to change, it is the
opposite of static.)
We did not even explain why Abdul and Bridget would have been at the original Nash equilibrium in the first place. Fortunately, the way we have derived
our best responses provides a way to fill in the necessary dynamic analysis.
Remember, when Abdul was selecting a best response he adopted a sim-
Figure 5.9: How players can get to the Nash
equilibrium: A dynamic analysis. Panel a.
shows the marginal costs and benefits of Abdul’s
fishing if Bridget fishes 12 hours. In panel b.
shows the dynamics of the choices in terms of the
fishermen’s marginal benefits and marginal costs.
The horizontal arrows show the direction Abdul
will move if he is initially at the base of the arrow.
The vertical arrows show the same for Bridget. The
inequalities involving marginal benefits and costs
(mb, mc) are the reason for the movement shown in
the arrows (which are called "vectors").
ple check list based on the marginal benefits of fishing more (mb) and the
marginal costs of fishing more (mc):
• if mb > mc, then fish more
• if mb < mc, then fish less
• if mc = mc don’t change how much you are fishing.
When we introduced this check list we focused on the last line, because that
is the equality that determines the utility-maximizing level of fishing for Abdul,
that is, it is a point on his best-response function.
The top two lines of the checklist tell Abdul what to change when he is not
fishing the optimal amount given what Bridget is doing, that is when he is ‘off’
his best-response function.
As Figure 5.9 shows, these first two lines of Abdul’s checklist tell us that
starting at any allocation (that is any combination of fishing hours of each of
them) in which direction he should move, shown by the arrows. The dynamic
analysis gives the following simple instruction: if you are not on your bestresponse function, move toward it.
Abdul’s arrows are green and horizontal (when he changes his fishing hours
M - C H E C K Abdul might adopt the instruction:
close half of the difference between the
hours I am now working and the hours
indicated by my best-response function,
given how many hours Bridget is now
working. For example, if Abdul were fishing
6 hours while Bridget fished 12 hours, he
would increase his hours by (12 6)/2 = 3
hours.
246
MICROECONOMICS
- DRAFT
he moves left or right). The same reasoning allows us to show the dynamic
arrows for Bridget, they are blue and horizontal, because when she changes
her hours that moves the allocation point up or down.
For example, in Figure 5.9 a, if Abdul is fishing 6 hours the marginal benefits
of fishing more exceed the costs (the bracketed term on the left). So in Figure
5.9 b, the arrows show that he will fish more. Similar reasoning (in reverse)
applies to the case where he is fishing, for example, 18 hours. The extent by
which the benefits differ from the costs depend on how much fishing Bridget is
doing. Figure a shows the case for when she is fishing 12 hours. You can also
work out how Bridget will adjust her hours if she is fishing more or less than
the amount indicated by her best response function.
You can see from the figure that unless the allocation is at point n one or both
of them will have an incentive to move (horizontally for Abdul, vertically for
Bridget) in ways that will lead them to the Nash equilibrium.
This explains why we would expect both Bridget and Abdul to be at (or very
close to) the Nash equilibrium. It also explains, if the Nash equilibrium shifted
because of some change in either a or b , why we would expect the two to
alter their fishing hours to move towards the new Nash equilibrium.
We now introduce a way that we can evaluate all of the possible equilibria of
this game by the standards of Pareto efficiency and fairness.
M-Note 5.6: Numerical Nash Equilibrium
In M-Note 5.3, we found the best responses for Abdul and Bridget given by Equations 5.18
and 5.19. Using the method we outlined above, we set Abdul’s Nash equilibrium hours of
fishing equal to Bridget’s BRF to find the Nash equilibrium level of fishing time:
Bridget’s BRF :
Collect terms
Multiply by
hAN
1
hA + hA
4
✓ ◆
5 A
h
4
4
5
=
15
1 AN
h
4
= 15
hAN
=
=
15
✓ ◆
4
15 = 12 = hBN
5
(5.27)
As a result, we see that each will fish 12 hours at the Nash equilibrium. Therefore they
each obtain the following Nash-equilibrium utility (by substituting hAN and hBN into their
utility functions):
uAN (hAN , hBN )
=
=
=
1 BN
1 AN 2
(h + hAN )
(h )
2
2
1
1
12(30
(12 + 12))
(12)2
2
2
216 72 = 144 = uBN
hAN (30
Each of them has a utility of 144 at the Nash equilibrium and the total welfare (sum of
utilities) is W N = uAN + uBN = 288.
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
Checkpoint 5.4: Storms and sustainability
Imagine that the external effect increased, as it would, for example, if greater
climate volatility produced storms that caused the two fishers fish in the same
limited part of the lake.
• Use the equation for the best responses of the two to redraw the figure. Why
do the fishermen best respond by fishing less?
• Use the equation for hAN and hBN to show that the Nash equilibrium level of
fishing will decline.
• Use what you have learned to explain how the best-response functions and
the Nash equilibrium would change if the fishermen jointly adopt a strategy
to let go of young fish to make the fish population more sustainable and
reduce the external effect they have on the other fisherman.
5.8 Evaluating outcomes: Participation constraints, Pareto improvements and Pareto-efficiency
Because the symmetrical interaction is just one of many possible rules of
the game that Bridget and Abdul might engage in, we need to go beyond
the Nash equilibrium for that game and think find a way to evaluate all of the
possible allocations that they might experience.
To do this, as in Chapter 4 we use the indifference maps of the two players
superimposed on the same set of outcomes. Recall that in the previous chapter every point in the Edgeworth box indicated an allocation composed of a
bundle of goods for Ayanda and another bundle of goods for Biko. We will see
that the same is true in this case if we plot an allocation as the pair of fishing
hours of the two, hA , hB . We start with Abdul’s preferences.
Because the utility of each depends on their own fishing time and the fishing
time of the other, that is because
Abdul’s utility:
Bridget’s utility:
uA (hA , hB )
B
A
B
u (h , h )
(5.28)
(5.29)
we can plot indifference curves with fishing time on each axis: Bridget’s fishing
time (hB ) on the vertical axis and Abdul’s fishing time (hA ) on the horizontal
axis.
We do this in panel a of Figure 5.10, where every point in the figure is a particular allocation of fishing times (hA , hB ). Using these allocations we can
calculate the utility that Abdul would experience were that allocation to occur.
On this basis we can calculate Abdul’s indifference curves based on in his
hours of fishing and Bridget’s hours of fishing. Abdul prefers curves labeled
with higher numbers, u3 > u2 > u1 . Notice two things about the indifference
curves:
247
248
MICROECONOMICS
- DRAFT
• The vertical dimension, or the effect of Bridget fishing more. Abdul’s preferred indifference curves are lower. This is because the less Bridget fishes
the better it is for Abdul.
• The horizontal dimension, or the effect of Abdul fishing more. If Bridget is
fishing at the "low" level indicated in the figure and supposing that Abdul
initially does not fish at all but considers fishing a little, he will start by
finding himself at successively higher indifference curves as he fishes
A
more, crossing the indifference curves labeled uA
1 , and then u2 and up to
uA3 . But if he spends too much time fishing he will then cross from uA3 back
A
down to indifference curve uA
2 and again go back down to u1 .
Another perspective on a best-response function
We can use the horizontal dimension of the figure to identify a point on Abdul’s
best-response function, associated with Bridget hypothetically fishing just
8 hours. We take this thought experiment as a constraint on Abdul’s utility
maximization. In the figure the green shaded area is his feasible set, and the
horizontal line is its frontier. Remember Abdul prefers indifference curves
that are lower down (indicating Bridget fishing less). The most preferred
indifference curve that is feasible is the one tangent to the constraint, at point
j, which is therefore a point in Abdul’s best-response function.
It may help to think of his indifference map as showing the contours of the
shoulder of a hill, and Abdul as walking along the frontier of the feasible set
towards point j trying out different amounts of time he might devote to fishing.
This is exactly what he did in Figure 5.6, comparing the marginal benefit
and marginal cost of fishing more. At first he is climbing – crossing contours
indicating ever-higher altitudes - higher utility. When he fishes 6 hours, he
A
achieves utility uA
k = 121, proceeding on to fish 8 hours, he achieves un = 144
, and finally fishing 13 hours, he achieves uAj = 169. At point j his path levels
off and if he continues to increase his fishing time he will descend to lower
altitudes – lower utility – once more.
Panel b of Figure 5.10 illustrates how two additional points on Abdul’s best
response-function are derived. The best-response function is constructed by
considering all of the possible levels of fishing that Bridget could hypothetically
do, and then reason as we did for point j.
Notice that Abdul’s best-response function intersects the indifference curves
where the indifference curves are flat. If the indifference curve is flat then
the mrs must be zero. In M-Note ?? we show why this must be true. In the
right panel you can see that as Bridget’s fishing time increases from 8 to 16
hours, Abdul’s fishing time declines from 13 to 11 hours. He identifies his bestresponse hours of fishing by finding the point on his best-response function
that corresponds to the number of hours Bridget fishes.
C O O R D I N AT I O N F A I L U R E S
k
uAk = 121
12
n
uAn
8
= 144
j
uAj = 169
B's hours,hB
B's hours,hB
16
hBHigh
& INSTITUTIONAL RESPONSES
A's best−response
function
= 16
k
hBN = 12
n
hBLow = 8
j
uAk
6
8
13
0
the indifference curves are flat?
The negative of the slope of the indifference curve is:
=
uAhA
uAyA
(5.30)
Abdul’s best-response function gives the values of hA and hB for which the derivative of
Abdul’s utility with respect to his fishing time is equal to zero or:
If uA
hA
=
a
b hB
15
18
(b) Abdul’s best-response function
M-Note 5.7: Why is the best-response function made up of points where
=
uAj
A's hours, hA
(a) Abdul making a choice constrained by Bridget’s hours
uAhA
uAn
11 12 13
A's hours, hA
mrsA (hA , yA )
249
(1 + 2b )hA = 0
0, then the numerator of Equation 5.30 is zero, so the slope of the indifference
curve is equal to zero, which means that it is flat.
Checkpoint 5.5: The marginal rate of substitution
Although each of the fishermen doesn’t "control" the hours of work that the other
does, we can still think in sensible ways about the marginal rate of substitution.
a. Consulting Figure ??, what is the sign of the marginal rate of substitution
going from left to right along the indifference curve leading up to point n on
uA2 (on the left-hand side of point n)? Why?
b. Continuing to consult Figure ??, what is the sign of the marginal rate of substitution going from left to right along the indifference curve on the right-hand
side of point n on uA
2 ? Why?
Fallback positions and the Pareto improving lens
We have said that rules of the game other than symmetrical interaction will
lead to different Nash equilibrium allocations. As long as the interaction
among the two is voluntary – there are no "offers you cannot refuse" – we
Figure 5.10: A new look at Abdul’s constrained
optimization problem for selecting his fishing
time depending on Bridget’s fishing time. .
To illustrate the construction of Abdul’s bestresponse function, in Panel a we consider Abdul’s
decision about how many hours to work, given
that Bridget has (hypothetically) decided to work
8 hours. The horizontal blue line is the constraint
on Abdul’s utility-maximizing process. In panel b,
we consider three hypothetical levels of Bridget’s
fishing time. The horizontal lines represent
Bridget’s fishing time at each of these levels, and
are the constraint on Abdul’s maximization process.
One of these horizontal lines is tangent to each of
Abdul’s indifference curves hB = 8 tangent to uAj
at point j, hB = 16 tangent to uAk at point k, and
hBN = 12 tangent to uAn at point n. Abdul’s entire
best response function is made of of points like j,
n, and k, for each of Bridget’s possible levels of
fishing hours.
250
MICROECONOMICS
- DRAFT
can limit the possible outcomes by thinking about the alternatives that the two
have should they decide not to fish at all. Any allocation in which they both
fish and that makes either of them (or both) worse off than how they would do
if they did not fish at all will not occur for the simple reason that they will not
fish if they could do better by not.
We have already introduced the idea that Abdul, if he does not fish at all,
will receive an income of yz possibly from family, friends or the government.
Suppose the same opportunity applies to Bridget. If they do not fish and
receive yz then their utility is just yz . This is their fallback position (like the
allocation z in the Edgeworth box of the previous chapter). But they only
receive their fallback if they do not fish, so the opportunity cost of fishing –
what they cannot have if they fish – is yz . In Figure 5.11 we show both Abdul’s
and Bridget’s indifference maps. We see from the numbering of the utility
labels on the curves, that Bridget’s indifference curves give greater values
the closer they are to the vertical axis (as Abdul’s did with the horizontal
axis).
One of their indifference curves is particularly important: it is labeled uA
z and
uBz . These two curves show all of the allocations hA , hB that yield, for Abdul
and Bridget respectively a level of utility equal to the utility of their fallback
position namely uz = yz . This is the participation constraint for each of them:
they will not participate in fishing unless they can do at least this well. Any
point between these indifference curves is a Pareto-improvement over their
fallback position: both are better off than their fallback option.
Any point between these indifference curves is a Pareto-improvement. The
Pareto-improvements are shown by the Pareto-improving lens shaded in
yellow.
The Pareto-efficient curve
There is another important curve in Figure 5.11: the purple solid and dashed
Pareto-efficient curve. We know that Pareto-efficiency requires that the fishermen’s indifference curves be tangent, that is, for their marginal rates of
substitution to equal. You can see two of these tangencies in the interior of
the Pareto improving lens. The other tangencies defining the Pareto-efficient
curve are not shown. The Pareto-efficient curve is made up of all points representing allocations for which:
mrsA =
uAhB
uAhA
=
uBhB
uBhA
= mrsB
(5.31)
Equation 5.31 shows the condition that the marginal rates of substitution must
be equal at Pareto-efficient outcomes, meaning that their indifference curves
are tangent.
C O O R D I N AT I O N F A I L U R E S
B's PC
21
uB3 = 155.8
B's hours, hB
18
uBz = 112
Pareto−efficient
curve
15
A's PC
12
9
uAz = 112
6
3
uA3 = 155.8
uA1 = 144
0
0
3
6
9
251
Figure 5.11: Abdul’s indifference curves and
Bridget’s indifference curves showing their
fallback levels of utility (their participation
constraints), uAz and uBz . At their fallback positions,
they are not fishing at all and receiving a payment
(in fish) of uAz and uBz from the government .
uB1 = 144
24
& INSTITUTIONAL RESPONSES
12
15
18
21
24
A's hours, hA
The figure clarifies the difference between Pareto improvements and Pareto
efficiency :
• The points on the purple Pareto-efficient curve that are indicated by a
dashed line outside the yellow Pareto improving lens are Pareto-efficient
but not Pareto improvements over the fallback no fishing option.
• The points in the yellow Pareto-improving lens that are not on the purple
Pareto-efficient curve are Pareto improvements but not Pareto efficient.
Checkpoint 5.6: Understanding the parameters
a. What would be the effect on the Pareto-improving lens if a increased?
b. What would be the effect on the Pareto-improving lens if b increased?
c. Why does b affect the person fishing even when no one else fishes? Why
does it make economic sense?
5.9 A Pareto inefficient Nash equilibrium
We return now to the symmetrical interaction between Bridget and Abdul, in
which the Nash equilibrium is the allocation at the intersection of their bestresponse functions. And we ask: is that allocation Pareto-efficient?
To answer, we combine two figures we have already introduced: Figure 5.11
showing the two fishermen’s indifference curves and Figure 5.8 showing their
252
MICROECONOMICS
- DRAFT
best-response functions. The combination of these figures results in Figure
5.12.
Figure 5.12 shows that the Nash Equilibrium is not Pareto-efficient: at the
Nash allocation (point n) the indifference curves of the two intersect rather
than being tangent. So allocation n cannot be Pareto-efficient.
How do we know that their indifference curves cannot be tangent at that point
that is, how do we know that
mrsA =
uAhB
uAhA
6=
uBhB
uBhA
= mrsB
(5.32)
The answer is that the Nash equilibrium is a point on both best-response
functions, defined by uB
= 0 for Bridget’s best response and uAhA = 0 for
hB
Abdul’s best response. At the best response each fisherman adjusts their
own fishing time to maximize utility so that these two terms will be zero. If we
substitute the zeroes for the marginal utilities in Equation 5.31, we find the
following:
• The first expression is now zero divided by uB
so the slope of Abdul’s
hA
indifference curve is zero; it is flat (as in Figure 5.12), and
• The second expression is now uA
divided by zero, so the slope of Abdul’s
hB
indifference curve is infinite; it is vertical (as in Figure 5.12 too)
A flat line cannot be tangent to a vertical line, so the condition for Pareto
efficiency is violated and the Nash equilibrium is not Pareto efficient.
A view from a Pareto inefficient status quo Nash equilibrium.
We now imagine Abdul and Bridget, fishing 12 hours each as indicated by
the Pareto-efficient Nash equilibrium. They realize they could both do better.
And they consider the options. They each might propose some different
allocation. To agree on an alternative level of fishing, the proposal would
have to implement a Pareto improvement. The Pareto improvement would
need to be over the Nash equilibrium, not over their no-fishing fallback option.
Remember that the Nash equilibrium is already better than their the fallback
positions.
With allocation n the new fallback for the agreement, we now have a new
yellow shaded Pareto-improving lens. There are two things to notice about
Pareto-improvements over the Nash allocation:
• both fishermen spend less time fishing and both are better off (have higher
utility than at the Nash equilibrium) and
• the new Pareto-improving lens is much smaller than the lens of Pareto
improvements over the no-fishing fallback option.
R E M I N D E R To understand Figure 5.12 it
will help to remember that for Abdul down
is better (his indifference curves have
higher utility the lower they are) because
the less Bridget fishes the better it is for
him. Similarly, Bridget is better off on the
indifference curves further to the left.
C O O R D I N AT I O N F A I L U R E S
uB3
uBn
B's hours, hB
Pareto−efficient
curve
●
hB* = 10
●
Figure 5.12: The Nash equilibrium and the
Pareto-improving lens. The Pareto-improving
fishing times (in which both fish less) are in the
pale yellow lens. Notice that Abdul’s indifference
curve at the Nash equilibrium is flat, and Bridget’s
at the same point is vertical (their marginal rates
of substitution are not equal). This being the case
there must be a Pareto-improving lens and the
Nash equilibrium cannot be Pareto-efficient.
n
i
A's best−response
function
uAz
uAn
253
uBz
B's best−response
15
function
hBN = 12
& INSTITUTIONAL RESPONSES
uA3
hA* = 10
hAN = 12
15
A's hours, hA
Checkpoint 5.7: Checking the marginal rate of substitution
Use the values of a = 30 and b = 12 and substitute them into the marginal rate
of substitution for Abdul and Bridget.
a. Confirm that when Abdul sets his mb = mc, the numerator is zero.
b. Also confirm that if you find the equivalent for Bridget, the mrsB = • (or it
is undefined) when Bridget sets her marginal benefits equal to her marginal
costs.
The reason why there exist allocations that are Pareto improvements over the
Nash is as follows.
• Reason 1: Each of them would benefit a lot if the other were to fish less
and
• Reason 2: at the Nash equilibrium each of them would experience very
little lost utility by themselves fishing a little less.
Reason 1 concerns each fishermen’s marginal utility with respect to the
other’s hours of fishing, and we can see that uA
< 0 and uBhA < 0 because
hB
each fisherman’s fishing time reduces the other’s productivity.
Concerning Reason 2, suppose that, at the Nash equilibrium level of the
fishing times, Bridget decided she would try to bribe Abdul to fish less. How
much would she have to give him to fish a tiny bit less? The answer is "almost
nothing" because at the Nash equilibrium, changes in his fishing time have no
254
MICROECONOMICS
- DRAFT
effect on his utility because the marginal benefits of fishing a little more or less
equal the marginal costs of fishing a little more or less (that is how he chose
that level of fishing to do).
So Abdul’s fishing a little less would not matter much to Abdul but it would
definitely benefit Bridget. A similar results is true for Bridget: Abdul could bribe
her to fish a little less for a tiny portion of his fish. This being the case if they
both could agree to fish less (and just forget about the bribes) they would both
be better off.
The conclusion is that Bridget and Abdul need not lament their sorry condition
at the Nash equilibrium. If a deal can be enforced – an agreement to limit
fishing, maybe along with a bribe – there’s a deal to be made that benefits
them both.
We turn now to considering changes in the rules of the game that might reduce fishing times, keeping in mind that we are thinking about not just two
people, but an entire community of people – perhaps the entire world’s population if we are considering coordination problems such as climate change or
the spread of epidemic diseases.
M-Note 5.8: The Nash equilibrium cannot be Pareto-efficient
To show that the Nash equilibrium is not Pareto-efficient we ask if they could agree each
to fish an arbitrarily small amount less would they both be better off. If the answer is "yes,"
then the NE cannot be Pareto-efficient. We know that uA
< 0 and uBhA < 0 each would be
hB
better off if the other fished less. We also know that uA
= 0 and uBhB = 0 because these
hA
equalities define Bridget’s and Abdul’s best-response functions, and the Nash Equilibrium
they are trying to improve on is a pair of strategies each of which is a best response to the
other.
So for any change dhA and dhB , representing an agreement to change their fishing time,
we can evaluate the change in each utility associated with change in the fishing times of
each.
duA = uAhA dhA + uAhB dhB
duB = uBhA dhA + uBhB dhB
Eliminating the terms equal to zero in the expressions above, namely those involving uA
hA
and uB
we have:
hB
duA = uAhB dhB < 0
duB = uBhA dhA < 0
or, rearranging
duA
= uAhB < 0
dhB
duB
= uBhA < 0
dhA
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
255
Both expressions are negative: the utility of each would be enhanced by an agreement to
fish a little less. The Nash equilibrium allocation of fishing times is not Pareto-efficient.
Checkpoint 5.8: Pareto-improvements
Use the numerical values of a = 30 and b = 12
a. How much utility would the players have if they both simultaneously reduced their hours of work from the Nash equilibrium values of 12 hours to 11
hours? Would they be better or worse off?
b. If one reduced their hours to 11 hours, what would the other’s best response
be? Would they actually reduce their hours to 11 or not?
5.10
A benchmark socially-optimal allocation
To provide a benchmark or standard against which we might evaluate the
various rules of the game that might improve on the Nash equilibrium of the
symmetric interaction above, we will reintroduce the Impartial Spectator, who
we relied on for the same purpose in Chapter 4. The Impartial Spectator
wishes to determine fishing time and distribute fish so as to maximize a social
welfare function, which, because she values the utilities of the two equally, is
just the sum of the utilities of the two:
Total social welfare
W
= Abdul’s utility + Bridget’s utility
= uA (hA , hB ) + uB (hA , hB )
(5.33)
She knows that the solution to this problem must be Pareto-efficient, because
if it were not, then one of the two could be made better off without worsening
the condition of the other, so this could not be the optimum for the Impartial
Observer, who values the well being of both. This means that the socially
optimal allocation must be somewhere along the Pareto-efficient curve in
Figure 5.12. But where?
A socially optimal allocation
To answer the question, we transform the view of the problem in Figure 5.12,
where the space in the figure is defined for hours of fishing, into a new graph,
Figure 5.13, in which presents the same information in terms of the utilities of
the two. The Pareto-efficient curve in Figure 5.12 appears in Figure 5.13 as
the dark green curve that is frontier of the feasible set of utilities.
Its slope is the marginal rate of transformation of Bridget’s utility into Abdul’s
utility. This provides the answer to the question: along the feasible frontier,
how much does Bridget’s utility have to fall in order for Abdul’s to increase by
one unit?
M - C H E C K In Chapter 4, we gave the Impartial Spectator Cobb-Douglas preferences
over the two players’ utilities. Here we have
the Impartial Spectator sum the utilities of
the two fishermen, as the Impartial Spectator
did in Chapter 3.
256
MICROECONOMICS
- DRAFT
W3 = 350
A's PC
B's utility, uB
225
W2 = 300
W1 = 250
i
150
112
B's PC
z
Utillity
possibilities
frontier
Feasible
utilities
112
150
A's utility, uA
225
You can see that when Bridget has almost all of the feasible utility then it does
not ‘cost’ Bridget much for Abdul to have a little more (the frontier is not very
steep); but the marginal rate of transformation rises (the curve steepens)
as Abdul gains more utility. The reason is that when Bridget has most of
the utility, she is working long hours (almost 15) and incurring a substantial
disutility of working time as a result. Fishing a little less would not reduce her
utility much, but for Abdul fishing a little more would substantially increase his
utility. So when Bridget is doing most of the fishing (and gaining most of the
utility) the opportunity cost of increasing Abdul’s utility (in terms of Bridget’s
forgone utility) is small.
The Impartial Spectator’s values are expressed by her indifference curves
(the blue lines), their slopes, her marginal rate of substitution, are a constant,
namely
1, because she values the utility of the two equally.
The point z represents the fallback utilities of the two (namely 112), and the
yellow shaded area is the set of feasible Pareto-improvements over this fallback position. You can see that the optimum point, i, is found where the highest feasible Impartial Spectators indifference curve is tangent to the utility
possibilities frontier (the frontier of the feasible set). So this is another case of
the familiar mrs = mrt rule, but now for the Impartial Spectator, rather than
Abdul or Bridget.
marginal rate of substitution
= marginal rate of transformation (5.34)
Figure 5.13: Feasible utilities, the utility possibility frontier, and the Impartial Spectator’s
iso-social welfare indifference curves. Here we
show the utility possibilities frontier and feasible
utilities for the Impartial Spectator. All points on
the frontier are Pareto-efficient. The points above
and to the right of the fishermen’s participation
constraints constitute the bargaining set, that is the
outcomes that are Pareto-superior to their fallback
options, uAz = uBz = 112. The Impartial Spectator’s
iso-social welfare indifference curves show her
equal valuation of the utility of the two and the negative of the slope of her iso-social welfare curves
is her marginal rate of substitution. The slope of
her iso-social welfare curves is -1 indicating that
she values the two utilities equally. The negative
of the slope of the utility possibilities frontier is the
marginal rate of transformation of Bridget utility into
Abdul’s. That is, it is the opportunity cost of Abdul
having more utility in terms of the utility that Bridget
forgoes as a result.
The impartial spectator will therefore choose
point i where mrs = mrt to maximize social welfare
given the constraint of the utility possibilities
frontier.
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
257
Rules that implement the social optimum
We know that acting on the basis of their best-response functions Abdul and
Bridget over-exploit the resource. They could both do better if they adopted a
different rule for deciding how much to fish. Before turning to institutions that
might implement such a new rule for their decisions, lets think about a rule
that would exactly implement point i in the figure.
To find the optimum, point i in Figure 5.13 the Impartial Spectator proposes
the following rules that, if followed, will maximize her social welfare function
Equation 5.33. We show in the M-Note 5.9 how these are derived.
hA = a
B
h =a
2b hB
2b hA
A
B
2b h
(5.35)
2b h
Focusing on the equation for Abdul, the socially optimal condition looks very
similar to Abdul’s best-response function (shown again below) when he was
maximizing his own utility, except for one big difference.
Marginal private costs
Abdul’s own optimality condition
h
A
= Marginal private benefits
b 2hA
= a
b hB
(5.36)
Comparing Equations 5.35 and 5.36 with the latter rearranged, or:
Marginal social costs
Abdul’s social optimality condition
A
A
h +bh
we see that the difference is that there is an extra
= Marginal private benefits
= a
2b hB
b hA (5.37)
b hB in the socially optimal
condition (Equation 5.35.) This is the negative external effect of Abdul’s fishing
on Bridget’s utility.
In Equation 5.37 we have moved this term to the left-hand side of the equation, adding it to the marginal private cost of fishing more (namely the disutility
of hours of fishing, hA ). So Equation 5.37, the condition for Abdul’s fishing
time to implement a social optimum says the following.
marginal private cost + marginal external cost = marginal private benefit(5.38)
The left-hand side is called the marginal social cost. The private cost
(marginal or average) is the cost that the decision-maker bears as a result
of some action that he or she takes. The social cost is the private cost plus
any costs imposed on others as negative external effects.
The Impartial Spectator reasons that if we are to implement an allocation
that values the utility of both equally, then Abdul should act is if he is taking
account of this cost – treating the costs he imposes on Bridget no differently
than his own disutility of labor – when deciding on how much to fish. As you
P RIVATE AND SOCIAL COST The private cost
(marginal or average) is the cost that the
decision-maker bears as a result of some
action that he or she takes. The social cost
is the private cost plus any costs imposed on
others as negative external effects.
258
MICROECONOMICS
- DRAFT
know from the previous chapter, this is called internalizing the negative external effect of his actions. We can think of these socially optimal responses in
the following way:
Optimal response = Own best response + Internalized cost to other (5.39)
Imposing the same condition on Bridget, the Impartial Spectator provides a
rule where each fisherman internalizes the negative external effect of their
hours of fishing on the other. As a result, we arrive at the levels of socially
B
optimal fishing time for Abdul and Bridget, denoted by as hA
i and hi :
Abdul’s socially optimal fishing time :
hAi
=
Bridget’s socially optimal fishing time :
hBi
=
a
1 + 4b
a
1 + 4b
(5.40)
(5.41)
Because b > 0, we see that each of the players’ Nash equilibrium levels of
fishing time are higher than the socially opimal levels:
hAN = hBN =
a
a
>
= hAi = hBi
1 + 3b
1 + 4b
These socially optimal levels of fishing time correspond to point i (for impartial)
in Figure 5.12 where Abdul and Bridget have the same fishing time (10 hours)
and the same level of utility.
The job description of the Impartial Spectator is not to figure out how this optimal allocation might be implemented ("way above my pay scale," she says).
We leave that to a second (also imaginary) person who we will introduce in
the next section.
M-Note 5.9: The Impartial Spectator’s Choice
We know that the Impartial Spectator has the following social welfare function, and we can
substitute the fishermen’s utility functions into W as follows:
Total Social Welfare
=
Abdul’s Utility + Bridget’s Utility
W
=
uA + uB
=
hA (a
=
ahA + ahB
b (hA + hB ) + hB (a
2b hA hB
b (hA + hB ))
1 A 2
(h )
2
1 B 2
(h )
2
1 B 2
(h )
2
1 A 2
(h )
2
(5.42)
Next, the social planner needs to find the social welfare maximum, or the optimal social
welfare. To do this, we partially differentiate W with respect to the hours of fishing of each
fisherman, hA and hB , as follows to find the first order conditions for the social welfare
C O O R D I N AT I O N F A I L U R E S
optimum:
WhA =
WhB =
∂W
∂ hA
=
hA
=
∂W
∂ hB
=
hB
=
a
2b hB
2b hA
a
b hB b hB
1 + 2b
a
2b hA
a
b hA b hA
1 + 2b
hA = 0
(5.43)
2b hB
hB = 0
(5.44)
Notice that Equations 5.50 and 5.44 look similar to the best-response functions we found
previously, but that each incorporates an additional b hB for Abdul in the numerator of
Equation 5.50 and b hA for Bridget as shown in the numerator of Equation 5.44. These
terms correspond to the cost of the external effect that each fisherman imposes on the
other.
M-Note 5.10: Numerical Choice of the Impartial Spectator
The Impartial Spectator has a social welfare function, W which is the sum of the two fishermen’s utilities. We substitute each fisherman’s utility function (Equations 5.11 and 5.12)
into the social welfare function and assume the parameter values of a = 30 and b = 12 .
W
=
uA + uB
=
hA (30
=
1 B
1 A 2
(h + hA ))
(h ) + hb (30
2
2
30hA + 30hB hA hB 2(hA )2 2(hB )2
1 A
(h + hB ))
2
1 b 2
(h )
2
(5.45)
The Impartial Spectator then needs to differentiate the social welfare function defined by
Equation 5.45 with respect to the two hours of work to determine how much each person
should work:
WhA =
WhB =
∂W
∂ hA
=
) hA
=
∂W
∂ hB
=
) hB
=
hB
30
hB
30
2
2hA
= 15
hA
30
hA
30
2
1 B
h
2
(5.46)
1 B
h
2
(5.47)
2hB
= 15
We can solve for the Impartial Spectator’s choice of work hours for Abdul and Bridget by
substituting Equation 5.47 for hB into Equation 5.46. Following the process of substitution
as usual, we find:
hAi
=
15
hAi
=
15
=
7.5
3 A
i
4
4
Multiply by hA
3 i
hAi
Collect terms
=
=
1
(15
2
1 A
h )
2 i
1
7.5 + hAi
4
4
30
(7.5) =
3
10
10 = hBi
Each of them will work 10 hours at the socially optimal and Pareto-efficient allocation of
B
hours that the Impartial Spectator made. At the allocation (hA
i , hi ), each of the fishermen
has a utility of 150, which is higher than they had at the Nash equilibrium (see Marshal
B
Memo 5.6) and the total social welfare is Wi = uA
i + ui = 150 + 150 = 300.
& INSTITUTIONAL RESPONSES
259
260
MICROECONOMICS
- DRAFT
Checkpoint 5.9: Changing things for the impartial spectator
Assume that a = 24 and b = 13 instead of the values you have used so far.
1. What would the number of hours be at the Nash equilibrium? How much
utility would each player have and what would the total utility be?
2. How many hours would the Impartial Spectator choose? How much utility
would each player have and what would the total utility be?
3. Sketch the best-response functions you found for (a) and sketch indifference
curves for the Nash equilibrium and the Impartial Spectator’s choice through
the relevant points.
Remedies: Preferences, power, and policy
Stories about two fictional people such as Abdul and Bridget and their textbook lake are light years away from real fishermen in Rhode Island or Australia. John Sorlien, the Rhode Island lobsterman who we quoted at the start
of the chapter, is in competition not with a single other fisherman, but with
hundreds. Unlike Abdul and Bridget (so far) real-world fishermen and lobstermen do sometimes cooperate to pursue common objectives (Sorlien headed
their association).
A friendly conversation and a handshake might be enough for Bridget and
Abdul. But how might such an agreement be arrived at, and how might it be
enforced in Rhode Island or Australia? An even greater challenge is how
to design and enforce similar agreements – for example to burn less carbon in order to mitigate climate change – that span not thousands of actors
but billions living under the jurisdiction of hundreds of independent governments.
But the parable of Abdul and Bridget has provided an important insight. Illuminating the basic source of coordination failures: the negative effect of
their own fishing on the other person (uA
and uB
, that is) is not part of the
hB
hA
utility-maximizing process by which each choose how much to fish.
Addressing these external effects is where institutions come in, meaning
changes in the rules of the game. There are three basic approaches whether
the common property resource that is being over exploited be a fish stocks
in a lake or the limited carbon emissions carrying capacity of earth’s atmosphere.
• Regulation of the exploitation of the resource by a government.
• Private ownership of the resource so that private incentives will deter overexploitation.
• Management of the resource through local interactions among the the
H I S TO RY In 1968, Garrett Hardin wrote
that “freedom in the commons means
ruin to all" and as a result he advocated –
“mutual coercion mutually agreed upon." But
Hardin’s pessimism overlooked the many
non-coercive ways that local communities
have prevented the tragedy.5
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
261
resource users.
These three approaches are sometimes referred to as states (meaning governments), markets, and communities, or similar terms.6
5.11
Government policies: Regulation and taxation
To underline the fact that we are not modeling actual governments but instead an ideal of what a well informed and set of public officials seeking to
implement better allocations might do, we will introduce a second hypothetical
character, the Mechanism Designer. He is tasked with finding policies to implement outcomes that would be recommended by the Impersonal Spectator
according to her values of efficiency and fairness. He is an economist, with
an engineering mentality, she is a philosopher and her job is to identify good,
better or best outcomes.
The mechanism designer’s job then is to implement the best that can be done.
The Mechanism Designer is the main character in the Chapter 16 which is
about public policy. You than think about him as an economist advising a
government about how to design and implement its policies.
A government has many options to address coordination failures such as
common property resource over-exploitation, including educating the public
about the costs and promoting basic research to find ways of making the
resource more sustainable. But we focus on just two:
• Fiat: Governments can sometimes order the implementation of an allocation – for example, reduced fishing – that they or the voters who elected
them prefer. A fiat is an order.
• Taxation: Governments regularly implement taxes as incentives: taxes
make activities more costly and can therefore, discourage these activities,
while still allowing each person or firm to choose how much of the activity
to engage in given the increased costs.
Fiat power
With respect to fiat, the government, if it knew all the relevant information,
B
B
could select hA = hA
i and h = hi to maximize total utility. The government
might implement this outcome by direct regulation, simply issuing a fishing
permit allowing each fisherman a certain number of hours. Any deviation from
the permitted hours would result in revocation of the permit and the fishermen
would have to revert to their no-fishing fallback positions.
Point i in Figure 5.12 is the Impartial Spectator’s efficient fishing time allocation that corresponds to what the government would choose. Assuming the
government had no reason to favor one fisherman over the other from the
F AC T C H E C K In response to the rapid
depletion of fish stocks in 1992 the Canadian
government simply banned fishing for cod in
the Grand Banks region of the North Atlantic.
Stocks have been recovering since then.7
262
MICROECONOMICS
- DRAFT
standpoint of fairness, a Pareto-efficient and equal distribution of fishing times
would be the fiat allocation.
Optimal taxes: Internalizing external effects
Rather than implementing the efficient fishing time plan by fiat, however, the
government might want to let the fishermen each decide how much to fish,
but to change their incentives in order to address the coordination failure. The
TAXES A tax is a charge the government
enforces on the production or purchase
of a good. A subsidy is a payment the
government makes to the producer or
purchaser, similar to a negative tax.
government would levy what is called a Pigouvian tax on fishing designed to
eliminate the discrepancy between the social and private marginal costs and
benefits of fishing.
The problem for the government is to select a tax rate on fishing time that as
an intended byproduct will motivate the fishers to implement an allocation
that maximizes total utility while at the same time maximizing their own utility.
This means bringing the fishermen’s private incentives (the utility function that
each maximizes) into alignment with the conditions laid out by the Impartial
Spectator.
The problem can be posed this way: find the tax rate that would transform
the utility functions of the two fishermen so that their individual best-response
functions are identical to those implied by the problem solved by the Impartial
Spectator: maximizing total utility and internalizing the costs of the negative
external effects.
To internalize the cost means to require each of them to pay (in taxes) for the
H I S TO RY Imposing taxes on particular
behaviors which the government wants
discourage because they impose negative
external effects on others – over fishing
fishing, smoking – the government is taking
an approach pioneered by the early 20th
century economists Alfred Marshall (1842–
1924) and A.C. Pigou (pee-GOO) (1877–
1959). In recognition of his contribution to
the field of what is called welfare economics,
these are sometimes called Pigouvian taxes.
reduction in the catch of the other that their additional fishing time imposes.
We know that each additional hour that Bridget fishes means that Abdul
catches b hA pounds less of fish. So to force Bridget to take account of this
negative external effect, she must be taxed at a rate of b hA for every hour she
fishes.
Bridget’s socially optimal tax rate depends on Abdul’s fishing time, because
the external effect of Bridget’s fishing on Abdul’s well-being depends on how
much Abdul fishes. If Abdul is not fishing at all, for example, there is no need
to tax Bridget’s fishing, because it has no external effect.
The tax that induces the fishermen to choose the socially optimal levels of
fishing time is just equal to the negative external effect they impose on others
at the Pareto-efficient levels of fishing time. A Pigouvian tax is a change in
the rules of the game that has the effect of internalizing the external effect
that is the cause of the coordination problem. The tax is an indirect form
of coordination: the fishermen as citizens elect a government which they
delegate to impose on them a set of incentives to overcome the over-fishing
problem.
E X A M P L E The "golden rule" is a common
ethical principle that people should treat
each other as they themselves would like
to be treated. A Pigouvian tax is designed
to accomplish the same result by imposing
on each decision maker the costs that their
decisions impose on others.
C O O R D I N AT I O N F A I L U R E S
M-Note 5.11: The best-response function of fishing with taxes
Equation 5.48 shows Bridget’s utility function when her fishing time is taxed at the rate of t
per hour fished:
uB (hA , hB , t )
=
hB (a
b (hA + hB ))
thB
(hB )2
2
(5.48)
To obtain the best-response function Bridget’s fishing conditional on the tax rate and
Abdul’s fishing time, we differentiate the equation above and set the result equal to zero:
uBhB
uB
hB
=
b hA
a
2b hB
hB = 0
t
We can rearrange this first order condition to say that (on the left-hand side of the equation below) the marginal benefits of fishing more (in fish caught) must be equal to (on the
right hand side) the marginal costs including the the disutility of additional fishing time plus
the taxes incurred by fishing more:
a
b hA
2b hB = t + hB
Re-arranging this to isolate hB so as to have a best-response function we have:
(1 + 2b )hB
=
hB (hA , t )
=
a
a
t
b hA
t b hA
1 + 2b
(5.49)
M-Note 5.12: A implementing the Impartial Spectator’s choice
We know from M-Note 5.9 that the first order condition for Bridget’s fishing time in the
social optimum allocation proposed by the Impartial Spectator is:
hB
=
a
b hA b hA
1 + 2b
(5.50)
The mechanism designer’s job is to find the tax rate per hour of Bridget’s fishing time that
will induce her to act as if that were her private (self-regarding) first order condition too.
We know from M-Note 5.11 that Bridget’s true best-response function including taking
account of the tax is the following:
hB (hA , t )
=
a
t b hA
1 + 2b
(5.51)
The question that the mechanism designer must now solve is: what is the level of the
tax rate t , that will make Equation 5.51 look like Equation 5.50 so that Bridget’s private
incentives will lead her to implement the Impartial Spectator’s social optimum.
Comparing the two equations you can see that setting the tax rate that Bridget pays
t B = b hA will make the two equations identical. So that is the optimal tax rate. The tax
rate for Abdul would, by the same reasoning be t A = b hB .
Then we can calculate the tax rate that Bridget pays at the Nash equilibrium. Because we
know that the optimal tax implements the social optimum recommended by the Impartial
Spectator, we substitute the value for hA
i into the expression for the tax rate so we have ,
t = b hAi or
Bridget’s tax rate
Abdul’s hours worked (Nash)
Bridget’s tax rate (Nash)
t
=
hAi
=
thB
=
b hAi
a
1 + 4b
ab
1 + 4b
& INSTITUTIONAL RESPONSES
263
264
MICROECONOMICS
- DRAFT
For the parameters we have introduced to illustrate the model with b = 0.5 and a = 30,
the tax rate at the Nash equilibrium of the game with optimal taxes is 5 pounds of fish per
hour of fishing time, meaning that every hour that Bridget fishes she must pay from her
catch a total of five pounds of fish. Given that she is fishing 10 hours at the equilibrium of
the game she pays 50 pounds of fish in taxes. Abdul pays the same amount.
We do not subtract this amount from the their utilities at the Nash equilibrium because
we assume that the total tax revenues are redistributed to the population equally without
regard for their fishing times.
Checkpoint 5.10: A partial tax
Assume a = 30 and b = 12 . Consider that a government implements a tax
where each person is charged t = 14 for an hour of work.
a. What is Abdul’s utility function incorporating this tax? What is Bridget’s utility
function with the tax?
b. What are their best responses to the tax policy?
c. What is the Nash equilibrium level of hours with t
= 14 ? Is the outcome
Pareto-efficient? If it is not Pareto-efficient, is it a Pareto-improvement over
the Nash equilibrium without the tax? Explain.
d. If t = 14 is insufficient to produce a Pareto-efficient outcome, what level of t
would produce the Pareto-efficient outcome? Why? Explain.
5.12
Private ownership: Permits and employment
But government policies are not the only change in the rules of the game
that might address the over-fishing coordination problem. Suppose that the
property rights over the lake are changed such that the lake is no longer a
common pool resource, but is privately owned. As a result, lake is no longer
non-excludable, as a resource that is both excludable and rival, the it is a
private good. The person who owns the lake, say Bridget, could exclude Abdul
entirely (remember that is what private property means). But as an owner she
now has bargaining power over Abdul, and may be able to do better by letting
him fish under conditions favorable to her.
How do these new rules of the game change the Nash equilibrium?
• Permits: Bridget might sell Abdul a fishing permit allowing him to catch not
more than a given amount of fish, setting the highest possible fee for the
permit consistent with Abdul being willing to fish under those terms.
• Employment: Bridget might offer Abdul an employment contract under
which Abdul would fish a given amount of time; the fish caught by Abdul
would be Bridget’s. Abdul’s compensation would be a wage (paid in the fish
caught by the two of them) which would be sufficient to offset the disutility
of Abdul’s fishing time and the opportunity cost of his fishing (and therefore
to satisfy Abdul’s participation constraint).
R E M I N D E R While a single owner will take
account of the costs of over-fishing and
restrict fishing accordingly, the owner may
also be a monopolist in selling the fish to
others, and will restrict fishing even more
than is socially optimal so as to sustain a
high price of fish. We do not include the
consumers of fish other than the owner and
those fishing on the lake. But we will include
them when we return to these effects of a
monopolist in Chapter 9.
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
Selling permits to fish
To understand what Bridget will do as the owner of the lake, let us return to
her utility function:
Bridget’s utility
uB (hA , hB )
= yB
1 B 2
(h )
2
(5.52)
In Equation 5.52, when Bridget was one of the two fisherman and could not
charge anyone to access the lake, her production of fish equalled her consumption of fish (yB ). Now, though, as she will be charging Abdul to access
the lake, we separate her production of fish, x(hA , hB ), from her consumption
of fish, yB . When she is the owner, her consumption of fish equals her own
production plus the fee she charges, F . Her utility therefore becomes:
Bridget’s
disutility
Bridget’s
production
Owner’s utility
z }| {
z }| {
1 B 2
B A B
A B
u (h , h , F ) = x(h , h ) + F
(h )
|
{z
} 2
(5.53)
Bridget’s
consumption
Equation 5.53 tells us that Bridget, as the owner, now has three variables to
determine not just one (her own fishing time) when she interacted with Abdul
in the symmetric game.
• her own fishing time, hB
• Abdul’s fishing time, hA
• the cost F to Abdul of the permit allowing him to fish hA hours.
Here, the fee for the permit will be an amount of fish that Abdul would transfer
to Bridget, but the idea easily extends to thinking about monetary payments
rather than payments in kind like this one, since fish in this scenario are effectively money.
Bridget, being self-regarding, will want to know: "what is the largest fee that I
can charge Abdul?" Remember Abdul has the option of not fishing at all and
and receiving a transfer of an amount yz of fish; this is his fallback position
with associated utility uA
z = yz . Agreeing to fish in Bridget’s lake means foregoing the fallback option, so uA
z = yz is the opportunity cost of fishing.
This is Abdul’s participation constraint, limiting how much Bridget can charge
for the permit: the fee plus the opportunity cost of fishing cannot be larger
than Abdul’s utility from fishing, uA (hA , hB ):
Abdul’s utility from fishing
Abdul’s participation constraint
A
A
B
u (h , h )
Permit fee + Foregone fallback
F + yz
Because Bridget would never consider charging Abdul less than she could,
we can assume that equation 5.54 will be satisfied as an equality and so
(5.54)
265
266
MICROECONOMICS
- DRAFT
(re-arranging the equation) we have:
F
Abdul’s participation constraint (PC)
= uA (hA , hB )
yz (5.55)
Bridget’s constrained optimization problem is now to vary hA and hB to maximize her utility from fishing plus the fee she charges Abdul, or:
Maximize her total utility:
uB (hA , hB , F )
subject to Abdul’s PC:
F
= x(hA , hB ) + F
= uA (hA , hB )
1 B 2
(h(5.56)
)
2
yz
(5.57)
We can use the expression for F that is equation 5.57 to replace the F in
equation 5.56, so that now Bridget’s objective is to chose hA and hB to:
Maximize:
uB (hA , hB )
= x(hA , hB )
1 B 2
(h ) + uA (hA , hB )
2
y(5.58)
z
B
Once we have found the hA
i and hi that maximize 5.58, we can insert those
values of hA and hB into equation 5.57 to determine F , the cost of the permit
to charge Abdul.
Comparing equation 5.58 with equation 5.33 we can see that Bridget maximizes the same quantity that the Impartial Spectator maximized, namely the
sum of the utilities of the two, except here yz is subtracted. But because yz
is a constant (112 pounds of fish in our numerical examples) the solution of
these two optimizing problems – the values of hA and hB chosen – must be
the same.
B
This means when Bridget is the owner the hours worked, hA
i and hi will be
equal, ten hours each in our numerical example – but the levels of utility
realized will be maximally unequal. Abdul will get exactly 112, his fallback
position, and Bridget will get 188 as shown by point b in Figure 5.14.
Because the allocation was determined by Bridget maximizing her utility subject to a constraint on Abdul’s level of utility (that is the participation constraint)
it has to be Pareto efficient, by definition. If Bridget as the owner implemented
a plan in which she reduced Abdul’s work hours and increased her own, she
would obtain an allocation on the utility possibilities frontier at point b0 . HowB
ever, if she implements the optimal number of hours (hA
i = hi = 10) at point
i and has Abdul pay her for the right to fish with a fee, then the total rents
available are 300. She will charge a fee such that Abdul will receive just a bit
more than his fallback, uA
z = 112 and she will get 300
112 = 188. This corresponds to a movement along the blue line with slope = 1 from point i to
point b. The blue line therefore indicates movements from point i to alternative
feasible trades when the fishermen are able to trade fish between them as
payments.
The economic reason why the result is Pareto-efficient follows directly from
the fact that Bridget knew in advance that she would capture all of the feasible
C O O R D I N AT I O N F A I L U R E S
slope = − 1
A's PC
225
b
B's utility, uB
188
178
bʹ
i
150
B's PC
112
z
112
150
A's utility, uA
225
rents. Given that fact she had every interest in making the total rents as large
as possible. Had she chosen a Pareto-inefficient outcome she would have
foregone the opportunity to make herself better of without making Abdul worse
off (than his participation constraint).
But why wouldn’t Bridget select hA = 0, and have exclusive access to the
lake? The reason is that the marginal cost of compensating Abdul’s fishing
time is very small when he is not fishing much, or at all. So it is to Bridget’s
advantage to let Abdul fish in the lake and pay her for the privilege, rather than
doing all the fishing herself.
Employing others to fish
Instead of issuing a permit, Bridget might hire Abdul to work for her. Employment differs from the permit system in that when Bridget employs Abdul, she
owns all of the fish caught by Abdul, but must devote some of this to paying a
wage w to Abdul sufficient to satisfy his participation constraint.
From our reasoning in the permit case we know that the participation constraint will be satisfied as an equality. This allows us to use the fact that the
total wage paid (w) must offset Abdul’s disutility of fishing time and the opportunity cost of fishing (namely his fallback option, yz that he gives up if he
fishes) or:
Abdul’s PC as an employee
w=
( hA ) 2
+ yz
2
(5.59)
& INSTITUTIONAL RESPONSES
267
Figure 5.14: Payments in fish takes the fishermen to allocations outside of the feasible set.
The blue line with slope -1 shows the allocations
of utility that are possible if the two fish at the
socially optimal times indicted by point i follow a
transfer of fish from on to the other. The slope is -1
because the opportunity cost of, say Bridget having
a kg more fish, is that Abdul has one kg less. If
Bridget as the owner implemented a plan in which
she reduced Abdul’s work hours and increased
her own, she would obtain an allocation on the
utility possibilities frontier at point b0 . However,
if she implements the optimal number of hours
(hAi = hBi = 10) at point i and has Abdul pay her
for the right to fish with a fee, then the total rents
available are 300. She will charge a fee such that
Abdul will receive just a bit more than his fallback,
uAz = 112 and she will get 300 112 = 188. This
corresponds to a movement along blue trade line
with slope = 1 from point i to point b.
268
MICROECONOMICS
- DRAFT
Bridget then must choose hA and hB to maximize her utility:
uB (hA , hB , w)
= xA (hA , hB ) + xB (hA , hB )
= A’s catch + B’s catch
1 B 2
(h )
2
B’s disutility
w
A’s wage(5.60)
Then using equation5.59 for Abdul’s wage, what Bridget maximizes when she
employs Abdul is:
uB (hA , hB ) = hB (a
b (hA + hB ))
(hB )2
+ hA (a b (hA + hB ))
2
( hA ) 2
2
yz
(5.61)
Equation 5.61 can be understood as follows:
• it is identical to what Bridget maximized in the permit case, namely equation 5.58, and
• identical to what the Impartial Spectator maximized, namely, equation 5.33
minus the constant yz .
In both the permit and the employment cases, the outcome is Pareto efficient, but Abdul gains only an amount equal to his disutility of fishing time
plus the opportunity cost of his fishing at all. The allocation proposed by the
Impartial Spectator and that implemented by Bridget as owner of the lake
does not differ in the fishing times of each, or the degree of exploitation of the
fishing stock. In this sense private ownership of the lake has addressed the
Pareto-inefficiency of the over-exploitation of the lake as a common property
resource.
The only difference is that in the private ownership case there is transfer of
rents (amounting to 88 pounds of fish in both cases) from the non owner to
the owner:
• In the permit case the transfer took the form of the fee for the permit to fish
(88) that Abdul paid to Bridget.
• In the employment case the transfer occurred because Bridget owned all
of the fish that Abdul caught (200), 88 pounds of which she retained for her
own consumption after paying him the wage (112).
This is general feature of social coordination problems. When one actor is sufficiently powerful to maximize their utility subject to the participation constraint
of other actors, a Pareto-efficient allocation will result, and the powerful actor
will get all of the economic rents. We have already seen this pattern in the
TIOLI power scenario in Chapter 4.
Checkpoint 5.11: Wages vs. Permits
Refer to M-Note ??. Assume the same values for the parameters of a and b .
C O O R D I N AT I O N F A I L U R E S
a. What wage would Bridget pay Abdul if Abdul’s fallback was zero?
b. What wage would Bridget pay Abdul if Abdul’s fallback was to have a job
working as an administrator for a utility of uA
z = 100 rather than working for
Bridget?
5.13
Community: Repeated interactions and altruism
Here and in previous chapters we have used two-person games to represent
economic interactions among a very large number of people. But some of our
interactions are with small numbers of people, for example, in our neighborhoods, families and workplaces, and even, in some cases, in exploiting a local
common property resource like a forest or fishery.
These small communities often address coordination problems in ways not
possible when the number of people interacting is very large. This is possible
because members of small communities:
• often have information about each other that is not available to governments or private owners who are not part of the community;
• interact repeatedly with each other repeatedly so that there are opportunities to retaliate against members who violate social norms or informal
agreements; and
• often care about each other, and these social preferences can reduce
conflicts of interest (as we saw in the previous chapter) and can provide the
basis for addressing coordination problems.
These characteristics of small communities give them capabilities in solving
coordination problems that are unavailable to purely government- or marketbased approaches. As we have seen in Chapter 2, public goods experiments
show that people are willing to punish fellow group members whose behaviors
violate norms, even when inflicting the punishment is costly to the punisher.
Lets see how a small community of fishing people – illustrated by Bridget
and Abdul – might address the over-exploitation of the common property
resource.
Repeated interactions
In the one-shot games we have introduced so far the strategies available
to the players are limited: select some amount of fishing hours. One way to
make the game more realistic is to let the interaction be repeated over possibly many periods with the same players. Then more complicated strategies
are possible, even if in every period there are just two actions one can take,
for example, fish ten hours or fish twelve hours. Importantly, strategies can
now be conditional on what the other player has done in previous play.
& INSTITUTIONAL RESPONSES
269
270
MICROECONOMICS
- DRAFT
Bina
140
d
140
144
●
156
●
b
144 c
(a) Stage game
One strategy might be play the strategy that would implement the social
optimum (fish 10 hours) in the first round and on the next and successive
rounds of the game, play whatever the other player played on the previous
round. This strategy is called "nice tit for tat": nice because it begins with a
strategy that could be mutually optimal, but "tit for tat" because it punishes the
other player if she takes the over-fishing option.
Consider a repeated game between Abdul and Bridget with the following
properties:
• Actions: In each period of the game, they may fish either 10 hours, the
socially optimal amount or 12 hours, the over fishing level at the Nash
equilibrium of the symmetric game in which they do not coordinate in any
way.
• Duration of the game: After every period that the game is played it is continued with some probability 0 < P < 1 We show in the Mathematics
Appendix that this means that the expected duration of the game in number
of periods is 1 1 P .
• Payoffs: In each period the payoffs are given in panel a of Figure 5.15. The
cell entries are from the analysis of their interaction we have carried out so
far, with the parameters used in our numerical examples. Payoffs for the
game are the sum of payoffs for each period the game is played.
• Strategies. Each may choose either to fish 12 hours in every period of the
game (called "Defect") or fish 10 hours in the first period of the game and
every subsequent period until the other fishes 12 hours, in which case fish
Grim Trigger
156
Defect
150
150 a
Grim Trigger
12 Hours
Aram
10 Hours
12 Hours
Aram
10 Hours
Bina
Defect
1500
1452
●
1500
1436
1436
1440
●
1452
1440
(b) Repeated Interaction
Figure 5.15: Repeated interactions can convert a
one-shot Prisoners’ Dilemma into an Assurance
Game allowing or coordination on a socially
optimal allocation. Panel a is the payoff matrix
for the one-shot (stage) game between Bridget
and Abdul, that is played once only. Inspection of
the payoffs shows that it is a Prisoners’ Dilemma.
You can confirm using the circle and dot method
introduced in Chapter 1 that each player fishing
12 hours is a Nash equilibrium (the circles and
dots show that this is also a dominant strategy
equilibrium. Panel b gives expected payoffs
for the game, if at the end of each period with
probability P = 0.9 the game is played again (with
the same payoffs per period as shown in panel a).
In this case the circles and dots indicate that the
repeated game has two Nash equilibria: the Nash
equilibrium of the one-shot game with payoffs to
each of 1440, and the socially optimal allocation
with payoffs 1500.
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
271
12 hours as long as the game lasts. This strategy is called "Grim trigger":
the term trigger is used because on act of defection by the other sets of a a
punishing defection by the actor, it is grim because the defections go on as
long as the game lasts.
How will this repeated game be played? We will assume that both players
are entirely self regarding. You can see that the single period payoff matrix
– called the stage game – has the structure of a prisoners’ dilemma. The repeated game will continue following each period with probability P, which for
concreteness we set as P = 0.9. This means that the expected duration of the
game is 1 10.9 = 10 periods. If the game were played among total strangers,
it is unlikely that it would be repeated with such a high probability. But if the
two players are neighbors or co-workers they are very likely to continue interacting.
The payoffs in the repeated game are derived as follows.
Playing Defect against Defect The payoff will be the payoff (144) of the mutual defect option of the stage game (fishing 12 hours) for as long as the
game continues (10 periods) so the payoff is 1440.
• Playing Grim Trigger against Grim Trigger : The payoff will be 150 (the
stage game payoff to fishing 10 hours) for as long as the game lasts, or
1500.
• Payoff to playing Grim Trigger against Defect: In the first period, the Grim
Trigger player fishes 10 hours, while the defector fishes 12 hours, and so
receives 140 that period; and then he defects in the next period (should
it occur) and until the game ends. The probability that the second period
happens is P and if it does it can be expected to continue for 10 more
periods so the payoffs from period two to the end of the game are 0.9 ⇥
10 ⇥ 144 or 1296 which, adding the first period’s payoffs totals 1436.
• Payoff to playing Defect against Grim Trigger : This is calculated exactly
is in the case immediately above. The defector gets 156 in the first period
and then the mutual defect payoff as long as the game lasts, totalling 1452.
Looking at panel b of Figure 5.15 you can see that mutual Defect is still a
Nash equilibrium in the repeated game: it is a best response to both Defect
and Grim Trigger, as the circles and dots show.
But in this case the repeated game is not a prisoners’ dilemma: the best
response to Grim Trigger is not Defect, but Grim Trigger itself. And so if the
two were to coordinate on fishing at the socially optimal level (10 hours) and
had decided to play Grim Trigger they would continue doing so until the game
ended.
This means the what was Prisoners’ Dilemmas if played as a one shot can
M - C H E C K We show in the appendix that the
expected duration of an interaction is the
inverse of the probability that at the end of a
period the interaction will be terminated.
272
MICROECONOMICS
- DRAFT
become an Assurance Game if played repeatedly if the game is repeated with
a sufficiently high probability. What this means is that entirely self-regarding
actors acting independently and without government regulation can escape
the prisoners’ dilemma.
You can confirm using the circle and dot method introduced in Chapter 1 that
each player fishing 12 hours is a Nash equilibrium (the circles and dots show
that this is also a dominant strategy equilibrium). Panel b gives expected
Bina
payoffs for the game, if at the end of each period with probability P = 0.9 the
game is played again (with the same payoffs per period as shown in panel a).
10 Hours 12 Hours
1440, and the socially optimal allocation with payoffs 1500.
M-Note 5.13: Cooperation without agreements in a repeated game
The key to how game repetition converts a Prisoners’ Dilemma stage game into an Assurance Game that can implement the socially optimal level of fishing is that Defect should
not be a best response to Grim Trigger. This requires that the payoff to playing Defect
against Grim Trigger should be less than the payoff to playing Grim Trigger against itself,
or (using he letters in the payoff matrix of panel a):
b+c
✓
P
1
P
◆
<
a
1
P
which, re-arranged to isolate the P gives us:
P>
b
b
a
c
This means that(in the one shot game) the probability that the game will be continued
after each round must be greater than the payoff advantage of defecting on a cooperator
(b
a) relative to the payoff advantage to coordinating on 10 hours rather than 12 hours
(b
c). This means that repeating the game is more likely to result in the Pareto superior
symmetric outcome (both fishing less) if
• the incentive to exploit the cooperation of the other is less
• the joint benefit of mutual restricting fishing hours is more and if
• the interaction will be repeated with high probability
For the payoffs in 5.15 a this condition is satisfied for any P > 0.5
Social preferences: Altruism
The fact that the community of fishermen is small means two important facts
about their context are likely to hold. First, people in small communities can
more easily access information about one another and engage in repeated
interactions as the basis for retaliation against those community members
who defect. Second, small communities are also often the basis of the people
having a concern about each others’ well-being. such as altruism, fairness
concerns, or reciprocity.8
To see how social preferences might help solve coordination failures, suppose
that in choosing an action each participant puts some weight on the utility of
Aram
equilibria: the Nash equilibrium of the one-shot game with payoffs to each of
12 Hours 10 Hours
In this case the circles and dots indicate that the repeated game has two Nash
a
a
b
d
c
d
b
c
Figure 5.16: Payoffs for the one-shot Prisoners’
Dilemma Game shown in Panel a of Figure 5.15
repeated here to accompany M-Note 5.13. For both
players: b > a > c > d .
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
273
the other as in Equation 5.62, so that Abdul and Bridget have social preferences like those we introduced in Chapters 3 and 4. Because we have used
uA and uB to refer to the utility each gets from fishing, we now introduce vA
and vB , which include their concern for the other’s well-being. Because they
are other-regarding, their evaluation of the outcomes they believe their actions
will produce are based on these social preference utility functions, vA and
vB .
Altruistic A:
vA (hA , hB )
= Own utility + l Other’s utility
= uA + l uB
Altruistic B:
vB (hA , hB )
= uB + l uA
We show in the M-Note that the best-responses maximizing these utility
functions are:
hB ( hA ) =
B’s best response :
(a
( 1 + l ) b hA )
1 + 2b
M-Note 5.14: The best-response function of altruistic people
Now, Abdul cares about Bridget. His utility function is:
vA (hA , hB )
=
uA + l uB
=
hA (a
b (hA + hB )
✓
1 A 2
(h ) + l hB (a
2
b (hA + hB ))
1 B 2
(h )
2
◆
To obtain his best response function, we differentiate Abdul’s utility functions with respect
to fishing time and set it equal to zero:
∂ vA
∂ hA
(1 + 2b )hA
hA
=
a
2b hA
=
a
(1 + l )b hB
a
( 1 + l ) b hB
1 + 2b
=
b hB
hA
l b hB = 0
Each takes account of a fraction, l , of the external effect that they have on
the other person. Concern for the well-being of other people thus might at
least partially substitute for a tax or government regulation in alleviating the
social coordination failure: when people care for each other they are willing to
internalize the external effect of the cost they impose on other people.
What level of concern for the other would implement the Pareto-efficient
outcome?
In order for altruism to implement the social welfare maximizing allocation
proposed by the Impartial Spectator, these altruism based best response
functions would have to to mimic those of the Impartial Spectator in Equations
5.40 and 5.41. This means that the (1 + l ) would have to take the value of
M - C H E C K Notice here that 0  l  1
rather than 0  l  12 in Chapters 3 and
4. This is because in those chapters the
utility functions are Cobb-Douglas whereas
in this chapter we are adding up the utility
functions to display altruism (making these
functions Cobb-Douglas would be very tough
mathematically!).
274
MICROECONOMICS
- DRAFT
2, meaning that we would have to have l = 1. Each fisherman would have
to be the perfect altruist that we defined in Chapter 2, caring as much for the
other person as she does for her self (namely l = 1). The difficulty of sustaining this level of altruism may suggest why most successful communities
do not rely entirely on good will, but supplement it with mutual monitoring and
punishment for transgression of norms.
Checkpoint 5.12
1. Refer to Equations 5.50 and 5.44. Why do we need l
fully Pareto-efficient outcome? Explain.
= 1 to implement a
2. What would it mean if l < 1? If people aren’t only altruistic towards each
other in a community, what else might they do to affect the utility others
receive if they break social norms by over-fishing?
3. What would it mean if l > 1. Would the outcome be Pareto-efficient? Why or
why not?
5.14
Application: Is inequality a problem or a solution?
Recall that there two standards that we use to evaluate policies to address
coordination failures:
• Does it result in a Pareto-improvement over the status quo, that is, does it
improve efficiency by making at least one of the participants better off and
none worse off?
• Is the resulting allocation more fair than the status quo, that is, are the rents
(improvements over the status quo) that the players receive fair?
In some cases the two objectives can be jointly realized; in others they are in
conflict.
The distribution of the economic rents resulting from coordination depend on
the particular transformation of the game which makes coordination possible.
Conflicts may arise about how best to address the coordination problems
that people face: some participants may prefer an inefficient solution to the
allocation problem because they get a larger share of the economic surplus at
a Pareto-inefficient outcome.
Unequal solutions to local social coordination problems are generally based
on the disproportionate wealth or power of one of the fishermen. It is easy to
see that if one of the fishermen has a much larger net than the others and so
can be assured of catching most of the fish, then his best response will approximate the allocation of a single owner of the lake. In this case, inequality
in wealth among the fishermen would lessen the coordination failure.
Important inequalities may exist even among otherwise identical fishermen. To
see this consider two possibilities:
E X A M P L E Even the most utopian, such
as the contemporary Amish or Hutterite
communities in the U.S., supplement
altruistic action with social sanctions and
monitoring of peers (which is easier to do in
small communities).
C O O R D I N AT I O N F A I L U R E S
uB2
uBn
uBf
B's BRF
n
hBN = 12
h = 11.8
uAn
●
B's hours, hB
BF
●
f
uAf
tB
●
Pareto−efficient
curve
uA3
A
hBtA = 9.4
t
& INSTITUTIONAL RESPONSES
275
Figure 5.17: First-mover advantage: Fishing
time-setting power. In the figure, Abdul is the
leader with fishing time-setting power (first-mover
power) and Bridget is the follower. Abdul doesn’t
have enough power to make a TIOLI offer, but
he can make a credible commitment of his own
fishing time such that Bridget will have to adapt
in choosing her fishing time to Abdul’s fishing
time. Abdul, when choosing his fishing time,
takes Bridget’s best-response function as his
incentive compatibility constraint. He maximizes
his utility subject to satisfying her incentive
compatibility constraint, finding the point at which
his indifference curve is tangent to her bestresponse function as occurs at point f where
Abdul exerts fishing time hAS and Bridget exerts
fishing time hBS and the two fishermen obtain
utilities uA2 = uAS and uB1 = uBS . This Stackelberg or
Leadership outcome is contrasted with the Nash
equilibrium outcome of the simultaneous interaction
where Bridget had higher utility uB2 = uBN and Abdul
had lower utility uA1 = uAN .
A's BRF
●
hAtA = 10.55
A
hAN = 12
hAF = 12.9
A's hours, h
1. Take-it-or-leave-it (TIOLI) power: When one player’s social position is
such that they have substantially more power than the other making a
TIOLI offer of both their own and the other’s fishing time, then they may
implement an allocation where they obtain all the economic rents and the
other remains on their participation constraint. As you already know from
the previous chapter, the outcome is Pareto-efficient.
2. First-mover advantage: When one player’s social position is such that
they can credibly commit to a fishing time such that the other must simply
respond, they obtain more of the economic rent than the other player, but
not as much as if they had TIOLI power. This is similar to the price-setting
power in Chapter 4. The outcome is Pareto-inefficient.
We start with first-mover advantage: the power to commit to one’s own fishing
time.
First-mover advantage: fishing time-setting power
Suppose that Abdul can announce a level of fishing time and commit to it in
such a way that Bridget understands that nothing she can do will alter Abdul’s
fishing activity. Bridget will then select her level of fishing to maximize her
utility given what Abdul has committed to. In this situation, Abdul is the firstmover and has fishing time-setting power similar to the price-setting power
(as in Chapter 4). Economists call Abdul in this situation the Stackelberg
leader.
H I S TO RY Heinrich von Stackelberg (19051946) used this model to represent pricesetting among duopolists (two firms in a
market), which is why this type of model is
named after him.
276
MICROECONOMICS
- DRAFT
The big difference between Abdul having fishing time-setting power and our
previous analysis is that the game is now sequential and the order of play
matters: who gets to go first is important.
How would Abdul decide what level of fishing time to commit to as the fishing
time-setter? As the first-mover, he will begin by determining what the secondmover will do in response to each of the his actions, and then select the action
that maximizes his own utility given the best-response function. The secondmover’s best-response function is the incentive compatibility constraint.
Abdul maximizing his utility subject to Bridget’s incentive compatibility constraint is a simple but important change in the assumed behavior of the fishermen: Abdul now recognizes and takes advantage of the fact that by choosing
various levels of fishing time he can affect the level of fishing time Bridget
chooses. Abdul’s behavior is strategic because it takes account of Bridget’s
R E M I N D E R In Chapter 4, when she had
Price-setting power, A maximized her utility
subject to B’s price-offer curve. B’s priceoffer curve was A’s incentive compatibility
constraint, which is exactly what a bestresponse function is: a best-response
function shows what action would be your
best response to the action (e.g. price or
fishing time level) the first mover commits to.
reaction to his action.
In this first-mover case, Abdul is constrained not by a given level of Bridget’s
utility, but by Bridget’s maximizing behavior as given by her best-response
function. As a result, the first-mover outcome will not be Pareto-efficient because Bridget’s indifference curve intersects Abdul’s indifference curves at
point f in Figure 5.17. Because the indifference curves intersect, the marginal
rates of substitution are not equal and therefore the allocation is Paretoinefficient. Similar to the analysis of price-setting power in Chapter 4, the
fishing time-setting outcome is not Pareto-efficient, because when Abdul maximizes utility subject to Bridget’s best-response function, he does not fully
internalize the external effect. Abdul’s first-mover advantage allows him to
improve his position by comparison to the Nash equilibrium, in this case at the
expense of Bridget whose outcome as second-mover is worse than the Nash
equilibrium.
Take-it-or-leave-it power
Let us switch roles now and consider what would happen if Bridget had more
power than Abdul. She has enough power to make Abdul a take-it-or-leave-it
offer, specifying not only how much she would fish, but how much Abdul is
to fish, too, along with the threat that should Abdul not accept the offer, then
Bridget would simply fish at the level of the Nash equilibrium of the simultaneous move game.
Because Abdul will refuse her offer if it is worse for him than the Nash outcome, Bridget must make an offer to Abdul that satisfies Abdul’s participation
constraint. If she does so, the outcome will be Pareto-efficient.
Referring to Figure 5.12, we can see the outcome that Bridget would implement if she had take-it-or-leave-it power over Abdul. In the case where Abdul’s
fallback position is the Nash equilibrium, then his participation constraint is
R E M I N D E R When either trader had TIOLI
power in Chapter 4, the exercise of power
led to a Pareto-efficient outcome.
C O O R D I N AT I O N F A I L U R E S
given by uA
n . As a result, Bridget would find the allocation on Abdul’s participation constraint at which she would maximize her utility. In Figure 5.12,
Bridget’s TIOLI offer is shown at point tB . At tB , her indifference curve uB
2
is tangent to Abdul’s indifference curve uB
n , so the allocation she chooses
for her TIOLI offer is Pareto-efficient, unlike when Abdul was the leader and
could set a fishing-time for Bridget. If Abdul had TIOLI power over Bridget, he
would implement the allocation at point tA and this allocation would also be
Pareto-efficient.
Summing up, our model shows that the effect on unequal power will always
benefit the more powerful and may, but need not, result in a Pareto-inefficient
outcome. Abdul’s fishing first-mover ability (his time-setting power), resulted in
gains for Abdul, losses for Bridget, and increased over-exploitation of the lake.
This resulted in greater inequality than the Nash equilibrium, but the outcome
was not Pareto-comparable to the Nash equilibrium (one does better and the
other worse at the Nash).
Positive effects on Pareto efficiency occurred when Abdul had take-it-or-leaveit power and the outcome was Pareto efficient, but probably regarded as unfair
by Bridget or an Impersonal Spectator. The model is hypothetical but the
problem is not.
Checkpoint 5.13: Numerical TIOLI power
Using the values of a = 30 and b = 12 :
1. Find the take-it-or-leave-it offer that Abdul would make to Bridget at point tA .
How many hours would Abdul fish? How many hours would Bridget fish?
2. What would Abdul’s utility be at the TIOLI offer? What woudl Bridget’s utility
be at the TIOLI offer?
Evidence from field studies
A field experiment among forest commons users in rural Colombia underlines
how inequality may be an impediment to achieving more satisfactory outcomes through coordination. Juan Camilo Cardenas implemented common
pool resource behavioral experiments among villagers who rely for their living on the exploitation of a nearby forest.9 So the subjects in the experiment
were in real life playing the same game that the experimenter invited them to
play.
In Cardenas’s game, the subjects choose to withdraw a number of tokens
from a common pool (these represented exploitation of the common property
resource), and after all subjects had taken their turn the tokens remaining
were multiplied by the experimenter and then distributed to the players, the
tokens then being exchanged for money. (This is similar to the Public Goods
Game experiment in Chapter 2 except that subjects decide how much to
& INSTITUTIONAL RESPONSES
277
278
MICROECONOMICS
- DRAFT
withdraw rather than how much to contribute to the pool). For an initial set of
rounds of the game, no communication was allowed. But in the final rounds of
the game, subjects were invited to converse for a few minutes before making
their decisions.
Cardenas expected that communication would reduce the level of withdrawals
from the common pool (as has been the case in similar experiments) despite
the fact that it does not alter the material incentives of the game. Communication was indeed effective among groups of subjects with relatively similar
wealth levels (measured by land, livestock and equipment ownership); their
levels of cooperation increased dramatically in the communication rounds
of the experiment. But this was not true of the groups in which there were
substantial differences in wealth among the subjects.
In one group, one of the wealthiest subjects tried in vain to persuade his fellow
participants (who in real life were his tenants and employees) to restrict their
withdrawals to the socially efficient amount, in order to maximize their total
payoff. But the wealthy subject’s advice fell on deaf ears.
“I did not believe Don Pedro,” one of the less well-off women in his group later
explained, “I never look him in the face.” She was right not to trust him: Don
Pedro (not his real name) had withdrawn the maximal amount despite his
contrary advice to the other players.
This is not an isolated example.
• A study of water management in 48 villages in the Indian state of Tamil
Nadu found lower levels of cooperation in villages with high levels of inequality in landholding. Moreover, lower levels of compliance were observed where the rules governing water supply were perceived to be chosen by the village elite.
• A similar study of 54 farmer-maintained irrigation systems in the Mexican
F AC T C H E C K In a recent study of participation in church, local service and political
groups, as well as other community organizations providing local public goods
by Alberto Alesina and Eliana La Ferrara
found that participation in these groups was
substantially higher where income is more
equally distributed, even when a host of
other possible influences are statistically
"held constant."10
state of Guanajuato found that inequality in land holding was associated
with lower levels of cooperative fishing time in the maintenance of the field
canals.11
In other cases, inequalities based on traditional hierarchical have made a
positive contribution.
• Another study of Mexican water management, for example, found that
increased mobility of rural residents undermined the relationships that had
been the foundation of a highly unequal but environmentally sustainable
system of resource management.12
• And in the port of Kayar, on the Petite Côte of Senegal, a cooperative
fishing time to limit the catch (to support higher prices, not to protect fishing
stocks) owed its success in part to the leadership of the wealthy local
F AC T C H E C K Social differences among
commons users affects outcomes in other
ways. The fishing agreement in Kayar was
threatened by conflicts between locals and
outsiders using differing technologies, and
other attempts to limit fishing failed due to
the indebtedness of fishermen to fish sellers
(who opposed the limits) and because the
wives of many of the fishermen were fish
sellers.
C O O R D I N AT I O N F A I L U R E S
traditional elite of elders.13
5.15
Over-exploitation of a non-excludable resource
We stated at the beginning of the chapter that we would illustrate the problem
of the common pool resource problem by the example of just two people. This
was despite the fact that, as a non-excludable resource, there would be no
limit on the number of people who could, if they wished, fish on the lake and
compete with Abdul and Bridget for the available stock.
We have so far studied just one of the two aspects of the coordination failure
resulting in over-fishing: the fact both Abdul and Bridget fished more hours
than was Pareto-efficient. They both could have been better off had they been
able to agree to fish less.
Now we introduce the second aspect of the problem: many more people
could fish the lake. Because the lake is a common pool resource, its nonexcludablity property of a means that there is open access.
How many people would use the lake? To answer the question we add more
context to the initial problem. Abdul and Bridget are part of a large community
of people who may make their livelihood fishing on the lake, or if not that, then
doing some other kind of work yielding a utility of uz (their fallback option).
People will decide to make their living fishing as long as the utility they gain
exceeds uz .
In Figure 5.18 using the same values for the parameters as in the other numerical examples in this chapter, we calculate the maximum utility that could
be attained by one person fishing alone, two (as in the case of Bridget and
Abdul), three, and so on up to 11. However many there are fishing, they will
receive the same utility in the Nash equilibrium because they are identical,
and we have so far assumed that none has any advantage in bargaining with
the others. The height of each bar is the utility attained.
When there are just two people fishing as in our previous examples involving
Abdul and Bridget each receives a utility of 144, as we found in M-Note 5.6.
The more people that fish in the lake the lower the utilities each of them receive will be. When there are ten they all have a utility of 21.3, barely greater
than their fallback options.
Now think about some other member of the community who is not currently
fishing but is thinking of doing so. Those fishing are doing better than the
fallback options. But if the 11th person decided to fish they would all receive
a utility of 18.4 (including the new fisherman). That is, they would all receive
less than their fallback options. So the eleventh person would decide not to
fish.
& INSTITUTIONAL RESPONSES
279
280
MICROECONOMICS
- DRAFT
240
220
200
180
Utility
160
140
120
100
80
fallback option = 20
60
40
Figure 5.18: The dynamics of over exploitation
of a common property (non-excludable)
resource. All of the people who might fish on the
lake have the same utility functions as Abdul and
Bridget with the values of a = 30 and b = 12 .
The height of the bar for a given number on the
x axis is the utility of each of the fishermen when
there are the indicated number fishing on the lake.
The fallback utility is uz = 20. You can see from
the figure that if the lake is a common property
resource, so that no fisher can be excluded, the
Nash equilibrium number fishing on the lake is
10 with each receiving a utility of 21.3. If the 11th
person fished on the lake, she – and all of the rest
of the fishermen there – would receive a utility of
18.37, that is, less than their fallback option). The
mathematics on which this figure is based are
shown in M-Note 5.15.
20
0
1
2
3
4
5
6
7
8
Number of people, n
9
10
11
Generalizing from this example, the Nash equilibrium number of people fishing
is the largest whole number of people fishing such that the utility that those
fishing receive is greater than or equal to the utility they would receive at their
fallback option.
As a result, the Nash equilibrium of this game is that we have:
• nN = 10 the number of people fishing and
• hN = 4.62 the number of hours each of them works.
We use the N superscript for each of these quantities because both are Nash
equilibria (but under different rules of the game):
• nN = 10 is a mutual best response because none of those fishing could
do better by not fishing, and none of those not fishing could do better by
fishing, and
• given that ten people are fishing, then hN = 4.62 is also a mutual best
response because for each person fishing this is a utility-maximizing choice
of hours, given the hours that everyone else is fishing.
The Nash equilibrium is Pareto-inefficient for two reasons: too many people
are fishing too many hours each. Just as was the case with Abdul and Bridget,
if each fished a little less they all would be better off.
And if fewer of them fished, all ten of them could be better off. Figure 5.18
shows that if 3 people fished they would each have a utility of 100. We call
people who are already doing an activity, such as fishing or owning a firm, incumbents. We therefore call the existing fishermen, the incumbents or incumbent fishermen. Suppose this was the case, and that the incumbents could
somehow agree to bribe the other 7 not to fish. Notice we have just changed
M AT H - C H E C K The equilibrium number of
people fishing must be a whole number
because the entry of a "fractional fisherman"
would not make much sense unless we
allowed people to split their day between
fishing and the fallback option, which we do
not.
C O O R D I N AT I O N F A I L U R E S
the rules of the game to allow the incumbent three to coordinate.
The incumbent three would have to give the other 7 potential fishermen an
amount of fish sufficient that each would be as well off as their fallback. This
amount would be uN
uz or 21.3
20, or 1.3 each. The total payments by
the three to the other seven would be 7 ⇥ 1.3 = 9.1, leaving each of the
three better off (each receiving 300 9.1)/3 = 97). If the incumbent three
increased the ‘bribe’ just a little bit then all 10 would be better of than at the
Nash equilibrium with open access.
M-Note 5.15: Nash equilibrium number of people fishing
Because access to the lake is open to all, the number fishing there will be the largest
whole number (because we cannot have fractions of people fishing) such that the utility
of those fishing is equal to than the fallback option (their utility if they are not fishing in the
lake), which is uz = 20. To determine this number, we first derive hN (n) the hours of fishing that each will do as a function of the numbers fishing, and use this result to determine
uN (hN (n)) the utility of those fishing as a function of how many there are.
To determine hn (n) we study the utility maximization problem of person 1:
max u1
h1
=
=
n
1 2
h
2 1
 hi )
h1 (a
b
h1 (a
b h1
i=1
b
n
 hi )
i=2
1 2
h
2 1
(5.62)
To find the hours of fishing that maximizes the utility of person 1 we differentiate equation 5.62 with respect to h1 , and set the result equal to zero. This gives us the first order
condition:
a
marginal benefit = a
n
b
2b h1
 hi
h1
=
0
b
 hi
=
h1 = marginal cost
i=2
2b h1
n
i=2
(5.63)
Rearranging Equation 5.63 we get person 1’s first order condition giving the utility maximizing amount of fishing time:
(1 + 2b )h1
h1
=
a
b
n
 hi
i=2
=
a
b Âni=2 hi
1 + 2b
(5.64)
All face the same first order condition so in the Nash equilibrium all fish the same amount
of hours: h1 = h2 = . . . = hN . Equation 5.64 becomes:
a
b (n 1)hN
1 + 2b
=
a
b (n
b hN
=
a
(1 + b + b n)hN
=
a
N
=
hN
=
(1 + 2b )hN
hN + 2b hN + b nhN
h
1)hN
a
1+b +bn
As in the rest of the chapter, we let a = 30 and b = 12 . You can verify that, if n = 10, then
& INSTITUTIONAL RESPONSES
281
282
MICROECONOMICS
- DRAFT
1 + b + b n = 6.5 and so hN (10) =
uN (10)
=
=
30
6.5
= 4.61. The utility of each fisher would be:
4.61 ⇥ (30
21.30
1
⇥ 10 ⇥ 4.61)
2
1
⇥ 4.612
2
(5.65)
If one more enters takes up fishing so that n
= 11, then the hours of fishing would be
hN (11) = 30
7 = 4.29.The new utility of each fisher would be now:
uN (11)
=
=
4.29 ⇥ (30
18.37
1
⇥ 11 ⇥ 4.29)
2
1
⇥ 4.292
2
(5.66)
But this ( n = 11) cannot be a Nash equilibrium, because everyone – including the new
entrant – would then be worse off than with the fallback option, uz
= 20. So the 11th
person would not enter (or if she did, others would leave). So the Nash equilibrium is
nN = 10. This is illustrated in Figure 5.18.
5.16
The rules of the game matter: Alternatives to over-exploitation
The new rules of the game allowing the incumbent three to bribe the others
is just a thought experiment demonstrating that the Nash equilibrium with
open access is not Pareto efficient. But commonly observed real-life rules of
the game – like inequalities in bargaining power, cooperative management
of the lake, or private ownership – could also address the over-fishing problem.
Checkpoint 5.14: Pareto efficiency and open access
a. Explain why open-access Nash equilibrium outcome with 10 fishermen is not
Pareto efficient. What alternative, if any, is Pareto superior to it?
b. Given your reasoning for a., do you think there are alternative outcomes that
are Pareto superior to, say, three fishermen bribing the other 10 not to fish?
Explain what the dynamics for the situations you describe would be? How
many fishermen? How many hours spent fishing? And so on.
TIOLI bargaining power
To see that the institutions governing the interactions among them matter,
think about the case in which one of the ten people fishing on the lake has the
power to make a take-it-or-leave-it offer to the rest. Here we have changed the
rules of the game by giving one of the fishermen TIOLI power (which allows
a kind of coordination). But the lake is still open access, so there are ten
fishermen there.
The one with bargaining power – suppose it is Abdul – can now say to the
others "each of you will fish x number of hours, and I will fish as many hours
as I wish." This is the “take it" part of the offer. The "leave it"" part is: "and if
you refuse, then I will return to fishing 4.61 hours." That is, return to the former
Nash equilibrium hours that occurred when there was no coordination among
the fishermen.
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
Open access, n=10
total hours: 46.15
283
A's Utility
Others (not A)
h = 4.62
h = 1.14
A has TIOLI power,n=10,
total hours: 21.56
h = 11.3
Cooperative, n=10,
total hours: 27.27
h = 2.73
h=6
A owns the lake, n=4,
total hours: 24
h=6
Cooperative, n=4,
total hours: 24
h=6
0
50
100
150
Utility
The other fishermen would know that without coordination the best they could
do is to all fish 4.61 hours, gaining a utility of 21.3. This is the others’ fallback
option to the TIOLI offer. If accepting Abdul’s offer made them worse off than
their fallback they would refuse, and just fish 4.61 hours. This is their participation constraint; if it is violated – so that they would receive a utility of less
than 21.3 – the others will not accept ("participate in") Abdul’s offer.
Abdul would know, therefore, that he needs to find the hours of all 10 of them
(his and the rest) that maximizes his utility subject to the participation constraint on the minimum utility the others can receive. Figure 5.19 shows the
offer Abdul would make, and the utility that he and the others would experience. The first row of the figure shows, from the previous figure, the result for
the unlimited access case without coordination.
When Abdul has TIOLI power the other fishermen work fewer hours (just
1.14 each, rather than 4.62 before), but get exactly what they had under the
uncoordinated open access case. This is so because that level of utility – 21.3
– is the participation constraint on what Abdul can offer them.
Abdul himself works 11.30 hours and enjoys utility equal to 153.2. Notice that
the total number of hours is reduced sharply compared to the uncoordinated
Nash equilibrium: from 46.15 to 21.56 hours. This reduction in total hours is
the reason why the others are able to fish less but still attain the same utility:
they catch more fish in an hour due to the lesser total hours of fishing.
200
250
300
Figure 5.19: The rules of the game: Noncooperation, bargaining power, and ownership.
The bars show the utility of the fishermen (it is
identical for all fishermen in the first and third row).
The numbers at the end of the bars show the hours
fished, where hA is the fishing hours of the owner
or person who has TIOLI power, h A is the fishing
hours of the non-owner or people who do not have
TIOLI power.
284
MICROECONOMICS
- DRAFT
A democratic fishing cooperative
An entirely different set of rules of the game – a democratic cooperative of
the fishermen – would implement a correspondingly different set of results.
Suppose that none of the ten fishermen has any bargaining power advantage
and that they jointly own the lake. They can decide jointly – democratically
by unanimous consent – on the same number of hours that each of them will
fish.
To figure this out they would think in the same way the Impartial Spectator
did when she maximized total utility. They will maximize the sum of their
utilities because this will also maximize the utility of each fisherman.
The
result is shown on line 3 of Figure 5.19. When there are 10 fishermen, they
would each fish 2.73 hours and attain utilities of 40.9 each. Because their
utility as coop members is now double their fallback option, others who are
experiencing the fallback utility of uz = 20 would wish to join the cooperative.
But it might be difficult to persuade the members to admit others, as this would
reduce the utility of the incumbent fishermen.
Their total fishing hours (27.3) is substantially greater than under the TIOLI
power of Abdul, and so is their total utility (409.1 compared to 344.9).
We can conclude two things from this last fact:
• Suppose Abdul still had TIOLI power. If the other fishermen could coordinate their actions, they could ’bribe’ Abdul to give up his bargaining
power and join their cooperative; they could have offered him the 153.2
that he received under his TIOLI power and still be better off dividing up
the rest of their utility (fish) amongst themselves. They would each receive
(409.1
153.2)/9 = 28.4, far better than the 21.3 they had when Abdul
had bargaining power. So shifting from Abdul holding TIOLI power to the
democratic cooperative is a Pareto improvement.
• The reason why this is the case is that under Abdul’s TIOLI power they
were as a group under -exploiting the fishing stocks. Abdul forced them to
do this because the less the other people fished the more fish he could
catch, and that was the only way he could increase his utility.
The reason why the cooperative’s decision results in a greater total utility than
the TIOLI case is that the members of the cooperative were pre-committed
to sharing the total utility equally. And so they each had an interest in making
total utility be as large as possible.
Things would have been very different if Abdul had had the power to take
some of the fish caught by the others (as in the employment and fee cases
we dealt with earlier). In this case he would have done the same as the cooperative. He would have implemented the fishing times that maximized total
utility. And then he would have taken fish from the others, leaving them just
C OOPERATIVE A cooperative is a business
organization or other association whose
members together own the assets of the
organization; they share the income resulting
from their activities and jointly determine how
the organization will be run (possibly through
the democratic election of a manager).
C O O R D I N AT I O N F A I L U R E S
enough fish so that they did not decide to stop fishing. The TIOLI allocation
was inefficient because Abdul’s bargaining power was limited.
The TIOLI case was not inefficient because Abdul had some bargaining
power. It was inefficient because he did not have enough power. As you will
suspect from the 2 person case studied earlier, the allocation would have
been Pareto efficient if Abdul had had all of the powers of a private owner of
the lake.
Private ownership
Under these new rules – private ownership of the lake – the lake is no longer
a common property resource because ownership means that Abdul can
exclude anyone he wishes from fishing. Abdul would make three decisions:
• How many other fishermen should I allow to fish in the lake? That is, what
is the total-utility-maximizing number of people who should fish the lake?
• How many hours should I allow them to fish?
• If I employ them, then what wage should I pay them? Or if I charge them a
fee for fishing, how large should the fee be?
You know how to answer the second and third questions from the case earlier
in the chapter when Abdul was the owner with just one other person Bridget
on the lake.
The first question is similar to that asked in Figure 5.18 but the answer is very
different. The number of people fishing the lake is no longer based on the
fishermen’s own decisions about where they can make a better livelihood.
This is not their decision to make. The owner determines the number of
others so as to maximize his utility.
How he would do this is shown on line 4 in Figure 5.20. Abdul will allow three
other fishermen to access the lake (so that means n = 4 including himself).
Going back to Figure 5.19 remember, there are now just four fishermen, not
10 as before. All four fishermen fish 6.0 hours with the owner receiving a
utility of 300 and each of the others receiving the same utility as their fallback
option, that is, 20.
The last line in Figure 5.19 shows what happens if Abdul is not the owner and
instead if all four of the fishermen were members of a democratic cooperative.
The members of the cooperative would implement exactly the same allocation
of work time as occurred under private ownership: 6 hours of work time each.
But the distribution of utility would be radically different, each of the four would
receive 90.
& INSTITUTIONAL RESPONSES
285
MICROECONOMICS
Owner's utility
286
- DRAFT
Figure 5.20: Utility of the owner when the lake
is privately owned. On the horizontal axis are
the total number of people fishing in the lake,
including the owner. So, for example, where n = 2
we have Abdul as the owner and there is one other
person, Bridget, the case we analysed earlier in
this chapter. The height of each bar is the utility
gained by the owner of the lake when he can both
determine how any people fish there and dictate
the terms under which they will work (as long
as they receive utilities superior to their fallback
position).
320
300
280
260
240
220
200
180
160
140
120
100
80
60
40
20
0
1
2
3
4
5
6
7
8
Number of people, n
9
10
11
We know that each working 6 hours is the allocation that maximizes total utility. The reason why private ownership of the lake implements this outcome
is that the owner is limited only by the participation constraints of the others,
and this is a constant (their fallback position of 20). So he implements an allocation to maximize the total utility, from which he must subtract the amounts
required to keep the three others "participating."
In sum, we can say the following:
• Open access leads to a Pareto inefficient over-fishing outcome in which all
the fishermen receive the same utility.
• TIOLI power implements to a highly unequal and Pareto inefficient underexploitation of the lake.
• Private ownership implements a Pareto-efficient and highly unequal outcome.
• A democratic cooperative implements a Pareto-efficient and equal outcome.
5.17
Conclusion
In practice, none of the approaches to addressing the common property
resource coordination problem could be expected to work perfectly as:
• no government is likely to have the information about the people’s preferences, production functions, and fishing times necessary to implement
Pareto-efficient fishing levels by fiat, or to design the optimal taxes that
would achieve the same result.
• private owners face some of the same problems due to lack of information
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
287
and moreover, while private ownership of a small lake or other common
property resource is conceivable, for may important common property
resources is infeasible (owning the oceans or the atmosphere for example)
or undesirable (think of the unaccountable power that a private owner of
such a vast common property resource could wield.)
• altruism towards close family and loved ones will lead us to take at least
some account of the effects of our actions on their well-being, but we are
less likely to know or care deeply care about how our actions affect total
strangers or even yet unborn generations who will benefit or suffer the
external effects of what we do.
The conclusion is not that the approaches to addressing coordination problems introduced here are ineffective. Each approach can contribute to making
economic outcomes more efficient and fair.
The models we introduced simplify the rules of the game that regulate how we
interact with each other in exploiting a common property resource, in contrast
to the vast diversity and complexity of rules that we observe. What the models
have done is not to represent the world as it is, but to identify key aspects of
how the world works to provide a lens for understanding them better.
Making connections
Social interactions and external effects: The interactions that economists
study include buying and selling in markets, but they also include nonmarket interactions, sometimes called ‘social interactions’ ("social" here
means simply non-market). The social interactions studied here include an
external effect: the fact that one person’s fishing reduces the catch of another, and this effect is not taken into account when each of the fishermen
decide how many hours to fish.
Public goods, common pool resources and club goods: All of these things
have the property such that each person’s actions have external effects on
others and in the absence of social preferences or policies that internalize
the external effects, these are not taken into account when people decide
how to act, resulting in outcomes in which some potential mutual gains
remain unexploited.
Policy: Government policies and institutions may be designed so that people
take account of external effects when they act. An example, is a tax on
fishing that imposes on each fisher the marginal costs that their fishing
imposes on the other fisherman, inducing each to choose their hours of
fishing as if they cared about these external effects as if it as the actor
herself rather than the other who bore the costs.
Property: Converting a common property resource into a privately owned
F AC T C H E C K Elinor Ostrom and her colleagues’ field research in different parts
of the world from Colombia to Switzerland
uncovered twenty-seven different local rules
for excluding others from access to common
property resources. These were based on
such things as residency, age, caste, clan,
skill level, continued use of the resource, use
of a particular technology, and so on.14
288
MICROECONOMICS
- DRAFT
resource may result in a Pareto-efficient outcome in which the owner
captures all of the potential mutual gains (rents).
Power: When a single person has all of the bargaining power and so can
make a binding take-it-or-leave-it-offer, he or she implements an outcome
in which there are no unrealized mutual gains, and all of these gains (rents)
go to the powerful person. Lesser forms of power – to commit to a particular fishing time, to which the other must respond, for example – advantage
the powerful and can result in inefficient outcomes.
Mutual benefits from coordination and conflicts over their distribution Policies
to address coordination failures differ in how the resulting rents are distributed; the resulting conflicts may make it difficult to agree on any policy.
Inequality: Differences in wealth, political connections and other sources of
power can be both a source and a consequence of inefficient and unfair
outcomes among people facing coordination problems. In some cases,
these differences can also mitigate the inefficiencies arising from coordination failures.
Models and relevance: Models, we wrote in Chapter 3, are like maps – a
simplified guide to the territory, not the territory itself. But the model of
social interactions introduced here, though quite abstract can be directly
applied to very concrete economic actions such as firms competing for
customers and, suitably extended, can illuminate global social interactions
and coordination problems such as climate change and the spread of
epidemic diseases.
Important Ideas
utility
disutility
external effect
common property resource problem
private property
coordination failure
rivalness
excludability
interdependence
Impartial Spectator
altruism
reciprocity
symmetrical interactions
asymmetrical interactions
social identification
TIOLI power
time-setting power
stackelberg leader
fiat power
government policy
decentralized implementation
tax
employment
wage
permit
participation constraint
fallback
incentive compatibility constraint
best-response function
binding participation constraint
social preferences solution
C O O R D I N AT I O N F A I L U R E S
& INSTITUTIONAL RESPONSES
Mathematical Notation
Notation
Definition
h
a
b
u()
v()
W
w
F
t
a, b, c, d
l
fishing times
parameter regulating the productivity of fishing times
external effect of fishing time on the other’s productivity
utility function
value function expressing an altruistic concern for the utility of another person
Impartial Spectator’s social welfare function
wage in the employment solution
permit fee in the permit solution
per unit tax in the government policy solution
payoffs in the repeated interactions game
extent of altruism (valuation of the other’s utility relative to one’s own)
Note on super- and subscripts: A and B: people; N: Nash Equilibria; i: Paretoefficient outcome; F: outcome with a first mover.
Discussion questions
See supplementary materials.
Problems
See supplementary materials.
Works cited
See Reference List.
289
Part II
Markets for Goods and
Services
293
Enter the Royal [Stock] Exchange of London, that place more respectable than
many a court; you will see there agents from all nations assembled for the utility
of mankind. There the Jew, the Mohammedan, the Christian deal with one
another as if they were of the same religion. There the Presbyterian confides
in the Anabaptist, and the Churchman depends on the Quaker’s word. ... They
give the name infidel only to those who go bankrupt.
Voltaire, 1734, Lettres philosophiques, Melanges (Paris, 1961) pp 17-18
When you hear the word "market" what other word do you think of? "Competition" probably is what came to mind. And you would be right to associate the
two words.
But you might have also come up with "cooperation". That is what impressed
Voltaire about the London stock market: mutually advantageous interactions, even among total strangers "from all nations assembled for the utility
of mankind." Markets allow us, each pursuing our private objectives, to work
together producing and distributing goods and services in a way that, while
far from perfect, is in many cases better than the alternatives. Markets accomplish an extraordinary result: unintended cooperation on a global scale,
although often with a highly unequal distribution of the benefits.
To better understand what markets do and how they work, begin with two
workaday facts: We acquire skills as we produce things and,for this and other
reasons, producing a lot of the same thing is often more effective in terms
of time and other inputs per unit than producing just one or a few of many
different things. This is called learning by doing.
Because of learning by doing and other advantages of large scale production
people do not typically produce the full range of goods and services on which
they live. Instead we specialize, some producing one good, others producing other goods, some working as welders others as mothers, teachers or
farmers.
There are huge advantages to this pattern of specialization – called the division of labor. Those who are naturally better at some task, or have learned to
be good at it by experience, or are in an environment in which it can be most
productively done can devote themselves entirely to what they are relatively
good at.
This is part of the explanation of why as a species we are so productive.
The limited number of species that have adopted a highly developed division
of labor – humans, ants and other social insects, for example – have out
competed other species. The total biomass of humans and the livestock we
have domesticated, for example, is estimated to be 23 times the weight of
all the other mammals on earth. And throughout most of human history the
biomass of ants – one of the most cooperative of species – has exceed that of
humans by a considerable amount.
H I S TO RY The Israeli historian Yuval
Noah Harari explains why it is our capacity to cooperate in flexible ways with
large numbers of other humans that
makes us unique among all the animals.
https://tinyurl.com/y3bpy4px
294
MICROECONOMICS
- DRAFT
But the division of labor poses a problem for society: once they are produced
by specialized labor, how are the goods and services to be distributed from
the producer to the final user. In the course of history this has happened in a
number of distinct ways from direct government requisitioning and distribution
as was done in the U.S. and many economies during the Second World War,
to gifts and voluntary sharing as we do in families today and was practiced
among even unrelated members of a community by our hunting and gathering
ancestors.
In a modern capitalist economy, the institutions that govern how the goods
and services are distributed from producer to user include markets, firms,
families and governments. In this second part of our book we study markets
and the actors who make up markets: the owners (and managers) of firms,
and other individuals (and the families of which they are a part).
To understand how markets facilitate specialization in chapter 6 we study the
production process and how the division of labor and the exchange of products can be advantageous to all concerned. Then to understand the workings
of markets we explain how individuals’ valuation of goods and services is
expressed in market demands (Chapter 7). Then, along with these market
demands, we explain how firms’ costs of production are expressed in their
owners’ and managers’ decisions about how much to produce and supply to
the market (Chapter 8).
We then study the process of competition among sellers and buyers, each
seeking to enlarge their share of the mutual gains made possible from the
division of labor and exchange. And we show how this so called rent-seeking
process affects the movement of prices and the quantities produced (Chapter
9).
Taking these four chapters as a whole poses a tension that can be expressed
by the following contradiction:
• The models and evidence on the advantages of large scale production
provide a reason why we specialize.
• Competitive markets are essential to the process of specialization can be
organized in ways that allow the mutual benefits of the division of labor to
be widely shared, as Voltaire said "for the utility of mankind."
• But the advantages of large scale production can also promote the emergence of giant firms and a winner take all process that appears to be
making markets less competitive.
Making market competition sustainable given the advantages of large scale
production will have to be addressed by public policy.
6
Production: Technology and Specialization
DOING ECONOMICS
The division of labor is limited by the extent of the market.
Adam Smith The Wealth of Nations
A technician glances quickly from one to the other of her three monitors and
around the huge room many other technicians do the same. Occasionally
a technician looks up at gigantic blue video screens on which news reports
flash, international weather reports display, and flight conditions stream live.
Twenty-four hours a day, translators stand ready to facilitate conversations in
28 languages.
What is this command center?
It’s the Production Integration Center that coordinates the global production
of the Boeing 787 Dreamliner four stories above the production floor at the
company’s plant in Everett, Washington, USA. There the super-jumbo airplanes are being assembled from components being flown in from around the
world: parts of the wing from Japan, wing tips from Korea, the center fuselage
and the horizontal stabilizer from Italy, passenger doors from France, cargo
doors from Sweden, landing gear from the U.K., and the list goes on and on.
Figure 6.2 shows where the components of the Dreamliner are produced. In
2015, Boeing contracted with over twenty-six thousand suppliers around the
world.1
Boeing selected Rolls Royce, Mitsubishi, Saab, Fuji and other companies to
design and build the components because they were – in Boeing’s estimation – simply the best companies to do the job, anywhere in the world. The
Japanese companies were global leaders in aircraft construction. The Italian
partner Alena had critical intellectual property rights (patents) that Boeing
would otherwise not have had access to.
This chapter will enable you to do the
following:
• Explain how learning-by-doing and
economies of scale are reasons for the
division of labor and specialization.
• Understand how markets allow specialization according to the principle of
comparative advantage.
• See how in the presence of economies of
scale and learning by doing an economy
can benefit by specializing, but also may
specialize in ways that perpetuate is
low income as a result of a poverty trap
similar to coordination failures studied
earlier.
• Manipulate some commonly used
production functions to study marginal
and average products of the labor and
capital goods and derive a production
possibilities frontier.
• Describe the main dimensions on
which technologies differ – the extent of
substitution among inputs, productivity,
factor intensity and economies of scale
– and how these are represented in
different production functions.
• Understand how a the owners of a firm
can determine a set of inputs and a way
of combining them to produce output
(a technique of production) that will
minimize the costs of a given level of
output.
• See that owners of a firm will try to
innovate to reduce the inputs required
to produce a given output and therefore
low costs and receive innovation rents (at
least until the competition catches up).
Boeing modified four 747-400 aircraft – renaming them "Dreamlifters" – to deliver the wings, body, and other parts of the plane to Everett where American
Figure 6.1: Boeing’s Production Integration
Center in Everett, Washington, USA.
296
MICROECONOMICS
- DRAFT
Figure 6.2: The different parts of the Boeing 787
aircraft built by different specialized partners
from all around the world.
machinists and others assembled the planes. Four stories above them the
engineers at the Production Integration Center kept minute-by-minute track of
the movement of the components around the world.
6.1 The division of labor, specialization and the market
Boeing’s globally integrated Dreamliner production process illustrates an
important economic idea: the division of labor. The division of labor is an
expression for the fact that people, organizations, or geographical regions
specialize in particular tasks or the production of a limited range of goods
or services. For Boeing, purchasing components of the Dreamliner from
hundreds of other specialized firms was more cost effective than producing
the entire plane in-house at their plant in Everettt, WA.
There are two consequences of specialization.
The first is increased productivity. The specialization allowed by the division of
labor increases productivity for three reasons:
• Comparative advantage: specialization enables people, firms, and regions
to focus on the tasks and products that they are comparatively good at (we
will take up comparative advantage below).
• Learning by doing: People learn better ways of working both through
the developing individual skills and discovering better ways to organize
production among members of a team. Figure 8.2 b provides a dramatic
example of learning by doing.
• Economies of scale: By allowing the production of a few things on a large
scale rather than many things on a small scale, specialization raises the
amount of output that is available for a given amount of inputs. Figure
8.6 presents some physical evidence for economies of scale based on
engineering studies.
A second consequence of specialization is the need for integration. The
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
advantages of the division of labor can only be realized if there are institutions
to coordinate the many distinct production activities that take place when
people specialize. The Boeing example illustrates the need for integration:
somebody has to put the parts together to produce a Dreamliner.
This can be summarized: production is specialized, but the use of goods
and services is generalized. Specializing in consumption is not biologically
sustainable. As a result, the goods and services, somehow have to get from
specialist producers to generalist users.
To grasp the scope of this problem imagine a 3D map of the world showing
the stocks of goods produced annually in each location. There would be a
hundred or so Dreamliners piled up in Everett Washington, billions of square
meters of cloth stacked in Bangladesh and other textile producing countries,
well over a billion barrels of oil piled over tiny Kuwait, mountains of computer
components and other consumer electronics rising from coastal China were
Dell is located, and so on.
Now imagine the same map, but showing where all of the goods are used.
The second map would be different from the first in two ways:
• it would be much flatter, the goods would have been spread around to the
entire population of the world and
• in any location there would be an assortment of a great many products, not
towering stacks of a single product.
The coordination of specialized producers and generalist users is accomplished by a set of institutions that differ in importance both over time and
across the economies of the world today. These include:
• Market exchanges: Selling the goods that specialized producers have
made provides the budget for purchasing the general market basket of
goods and services on which we live.
• Government acquisition and provision: Publicly provided services are
based on the integration of the specialized producers goods and services
to provide education, security and other government-provided services to
generalist users of these services.
• Families and other face-to-face communities: Typically families exhibit a
division of labor by age and gender: adult women, for example, biologically
producing children and spending disproportionate time on raising them and
caring for other family members (for example preparing meals). The goods
produced and tasks performed by adult men and women and by children
are shared within the family or some other larger consumption unit.
These three ways of coordinating the division of labor have in common that
they are like a two-sided platform that connects specialist producers with gen-
297
298
MICROECONOMICS
- DRAFT
0.72
Slope of ray from origin
equals Average Product
Total product, x
Total product, x
x=
Total Product
1
x = (l)2
50
0.97
Slope of tangent line
equals Marginal Product
0.55
4
2
6
Marginal product
1
mp(l) = (l)
25
0.24
0.16
0.12
Average product
1
ap(l) = (l)
50
0.08
0.04
2
4
Average & marginal product, ap, mp
Average & marginal product, ap, mp
Slope of ray from
origin equals
Average Product
Slope of tangent line
equals Marginal Product
2
(l)2
Total product
1
x = ln(1 + l)
2
0.8
0.32
0.08
1
50
4
6
0.27
Average product
1
(ln(l + 1))
2l
0.2
0.17
0.16
ap(l) =
Marginal product
1
mp(l) =
2(l + 1)
0.1
0.07
6
Hours of labor, l
(a) Economies of scale
eralist users of goods and services. – like Air BnB that matches home owners
to people looking for a place to stay, or Tinder, a dating app. This chapter’s
head quote by Adam Smith tells us that markets play a critical role in allowing
the division of labor to expand to global proportions, leading to ever greater
specialization. Here, in Part III of this book and also Chapter 14 we explain
how markets work to coordinate the division of labor. We begin with the production process, and some aspects of it that favor specialization.
6.2 Production functions with a single input
In Chapter 3, we used information on the way that Aisha’s study time translated into her learning to ask how much time she will choose to study. In
Chapter 5 information about the relationship between fishing time and the
amount of fish caught was a key idea in posing and then addressing the common property resource coordination problem. In both cases we were using, as
you recall from Chapter 5, production functions.
To better understand the division of labor and specialization we now need to
look more carefully at the properties of production functions. Think about another person, Alex, who has to choose how much of his time to spend produc-
2
4
6
Hours of labor, l
(b) Diseconomies of scale
Figure 6.3: Production functions with
economies and diseconomies of scale.
With economies of scale (Figure a) doubling
labor input more than doubles output, as can be
seen by going from 2 hours of labor to 4 hours.
Average product and marginal product increase
with labor input as you can see from the slope of
the production function (the marginal product),
which steepens as the labor input increases and
the slope of the ray from the origin to a point on
the production function (the average product),
which also steepens. The average and marginal
products given by these slopes in the upper figure
are shown in the lower figure of panel a. With
diseconomies of scale (panel b), the opposite
occurs: the average and marginal products of labor
are both decreasing.
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
299
ing one or more goods. Alex can devote more or less labor (l ) to production
and he can observe how his output (x) varies as he changes the number of
hours he works. Alex may use a computer and a desk or a given plot of land
and farming equipment or other inputs, but for now we assume that the only
input is his labor time.
The relationship between the input of his labor and the output of the goods is
described in a production function – x = f (l ) – taking the form:
x(l ) = ql a
(6.1)
The exponent a measures the responsiveness of output to a change in the
R E M I N D E R A production function is a
mathematical description of the relationship
between the quantity of inputs devoted to
production on the one hand (the arguments
of the function) and the maximum quantity of
output.
level of labor. The positive constant q measures the overall productivity of the
production process, which will be greater the more skilled or hard-working
Alex is. The top panel in Figure 6.3 a illustrates this production function (equa1
tion 6.1) with q = 10
and a = 2. The top panel in 6.3 b shows a different
production function: x = 12 ln(1 + l ).
In the top portion of both panels more hours of labor result in more output, but
the panels differ in how much output is obtained for given inputs.
E CONOMIES AND DISECONOMIES OF SCALE
When production exhibits economies of
scale, increasing inputs by a factor more
than proportionally increases output; with
diseconomies of scale, increasing inputs by
a factor less than proportionally increases
output.
• Economies of scale (6.3 a): when Alex doubles all of the inputs – in this
case that means just his labor input – the output more than doubles.
• Diseconomies of scale (6.3 b): when Alex doubles his labor input (assumed to be the only input), the output less than doubles.
The term constant returns to scale, not shown in the figure, refers to the
case where when inputs double output doubles, so the production function is
just a straight line as shown in the lower right panel of Figure 6.4.
In the lower figures of both panels we show two important statistics describing
aspects of the two production functions in the above figures. The ratio of the
amount of output to the amount of the input involved in producing it is the
average product of that input (also called average productivity).
Average product is measured by the slope of the line from the origin (called a
AVERAGE PRODUCT The average product
of labor is the ratio of the output to the labor
input.
ray) to a point on the production function. The bottom panel in figures a and
b show the average product of labor associated with the production function
shown at the top of those figures. When production exhibits economies of
scale, average product increases as the scale of production increases through
an increase in inputs.
The ratio of the increase of output to an increase in labor input is the marginal
product of labor (also called marginal productivity).
M-Note 6.1 summarizes the different cases when the output is x and the only
input is labor, l .
M ARGINAL PRODUCT The marginal product
of labor is the ratio of the change in total
output to a small change in input.
300
MICROECONOMICS
- DRAFT
Checkpoint 6.1: Production and labor inputs
Consider a production function: x(l ) = 10l 0.5 :
a. Sketch the production function.
b. Calculate ap(l ) and mp(l ). Sketch them.
c. Does the production function exhibit economies or diseconomies of scale?
M-Note 6.1: The average and marginal product
Here we summarize the concepts of total product, average product and marginal product. Because the marginal product is the slope of the production function, it is also the
derivative of the production function with respect to a particular input, e.g. for a production
function using only labor, xl =
d f (l )
dl .
If there is just a single input, labor, and the marginal product of labor is greater than the
average product of labor then the average product must be increasing as more labor is
used. We will use equation 6.1 to illustrate why this must be the case.
We start calculating the average product ap and marginal product mp of the production
function x(l ) = ql a :
Average product:
ap
=
Marginal product:
mp
=
x(l )
ql a
=
= ql a
l
l
dx(l )
= aql a 1
dl
1
(6.2)
(6.3)
To analyze how the average product changes as more labor is put into production, we
calculate the derivative of the ap function with respect to labor hours:
dap(l )
dl
=
(a
1)ql a
2
If a > 1, the expression above is positive, which means that the ap increases with more
labor. If a < 1, it is negative: the ap is reduced if we add labor.
In summary:
• If a > 1, then mp > ap, and so
• If a < 1, then mp < ap
dap(l )
dl
dap(l )
and so dl
> 0 and
<0
6.3 Economies of scale and the feasible production set
Suppose that Alex can spend his time fishing and making shirts in some
combination, including complete specialization (spending all of his time on one
or the other). He prefers to have more of both shirts and fish: both are goods.
He needs at least some of each to survive. His labor time, as in Chapter 5 is
a "bad" but we will set aside his choice of total hours of work by saying that
he can work any amount up to ten hours a day, and that given how productive
his labor is and how much he values the goods, he will choose to work the full
10 hours.
As a result, the more time Alex devotes to producing one good,
the more of that good he will have, but because Alex’s time is limited, the
less he can produce of the other. Therefore the opportunity cost of more fish
P RODUCTION POSSIBILITIES FRONTIER
(PPF) The production possibilities frontier
for two goods shows the maximum feasible
amount of one good that can be produced
given the output of the other. The production
possibilities frontier is the boundary of the
producer’s feasible set and is an alternative
name for the feasible frontier when we study
on production.
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
301
is the amount of shirts he will have to forego if he shifts his work time from
shirt-making to fishing. In order to pose the question – how much time will he
spend on each? – we need two pieces of information:
• the feasible set of combinations of fish and shirt amounts that are available
to him, given his labor time and the production functions at his disposal;
and
• the indifference map representing his valuation of each of the combinations
of the two goods.
We derive the feasible set in this section and introduce the indifference map in
the next.
We will assume that there are economies of scale in shirt making (as is common in manufacturing processes) but constant returns to scale in fishing. This
means that the fish production function is linear: it is a straight line. As a result
both the average and marginal products of labor are constant and equal to
each other and also equal to the slope of the blue line through points f an c in
Figure 6.4. In in Figure 6.4 we derive his feasible set of fish and shirts based
on
• the total amount of time he will work (shown in the lower left quadrant of
the figure); and
• the productions functions for fish and shirts (shown in the lower right and
upper left quadrants, respectively).
In the figure the horizontal axis to the left of the origin represents positive
amounts of labor devoted to making shirts, and the vertical axis below the
origin is positive amounts of labor devoted to fishing.
Alex needs to make a choice between three different ways to allocate his time
in production to two types of output (fish and shirts):
a. Allocate 10 hours of work to producing only shirts
b. Allocate some hours of work to producing shirts and some to fish, and
c. Allocate 10 hours of work to producing only fish
Option a) is shown as point a in the figure where Alex dedicates all of his 10
hours of work time to producing shirts. We extend a line up to his production
function for shirts and notice that 10 hours of labor results in Alex producing
50 shirts.
We extend a dashed line to the y-axis to see what this would correspond to on
the production possibilities frontier and see that Alex would produce 50 shirts
and no fish as a result of dedicating all his labor to shirts. We could follow the
M - C H E C K Remember, the average product
is the slope of a ray from the origin to a point
on the production function, and the marginal
product is the slope of the production
function. So if the production function is
just a straight line, both of these are equal
and do not vary as more labor is devoted to
production.
MICROECONOMICS
- DRAFT
a
10
10 hrs of labor
for shirts produces
10 shirts
Shirts, y
302
Shirt Production
1
y = (ls)2
10
e
10
2.5
Feasible
set of
outputs
Labor for Shirts, ls
5
b
5
2.5
Fish, x
Figure 6.4: Deriving the production possibilities
frontier with economies of scale. The lower left
quadrant shows the constraint: a given amount
of labor is available. The upper left and lower
right show how the available labor can produce
shirts and fish respectively. The upper left is an
economies of scale production function similar to
the panel a in Figure 6.3, but just rotated clockwise
280 degrees. The lower right production function
has constant returns to scale. Points d, e, f, and b
illustrate the production of shirts and fish that are
possible if the labor time is divided equally among
the two sectors. To check that you understand
how the figure works, find the point on the feasible
frontier associated with devoting 8 hours to fishing
and 2 to shirt-making, (trace out the new points, d’,
e’, f’, and b’).
Feasible
set of
labor hours
d
5
Constraint
on total
labor hours
S
l + lF ≤ 10
10
Labor for Fish, lf
lS = 5, lF = 5
f
Fish Production
1
x = (lf)
2
c
10 hrs of labor
for fishing produces
5 kgs of fish
same process for option c) corresponding to point c in the figure. He would
produce 5 fish and no shirts by dedicating all his time to fishing.
Option b) (that is, point b) on the other hand, shows what Alex would produce
by dedicating half his time to shirts and half to fishing. Because production in
fishing is linear, if he dedicates 5 hours, he simply gets half of what he would
have produced at 10 hours (2.5 fish). But, because there are economies of
scale in shirt production, from dedicating half his time to shirts, he only gets a
quarter of the output relative to 10 hours of labor for shirts (12.5 shirts vs. 50
shirts)
The top-right quadrant of Figure 6.4 illustrates economies of scale. The result is that Alex’s production possibilities frontier is bowed inward toward the
origin. This reflects the fact that with economies of scale in one or both production functions, dividing your work time between the production of both is
not as good (it is closer to the origin) than devoting all your time to just one or
the other.
The (negative of the) slope of the production possibilities frontier shows the
opportunity cost of acquiring more fish by shifting labor from shirt-making to
fishing, in terms of the amount of shirts that must be foregone as a result.
With economies of scale, as Alex shifts his labor input from producing clothing
to producing fish, mrt (x, y) declines, so he gives up smaller and smaller
amounts of clothing to get larger and larger amounts of food.
This reflects the fact that with economies of scale the marginal product of
M AT H N OT E We describe the production
possibilities frontier with economies of scale
as convex toward the origin, that is, it is
bowed in to the origin.
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
303
his labor decreases the less labor he devotes to production of a good. This
means that when he is doing little shirt-making, his marginal productivity in
that activity is low, so doing a little less (so as to allow him to do more fishing)
does not result in a large reduction in shirts produced.
6.4 Economies of scale, specialization and exchange
In Figure 6.5 we combine the feasible set from the Figure 6.4 with an indifference map represented by the three numbered indifference curves. Recall
that, as indicated by the numbering of the curves, farther away from the origin
is better in Alex’s evaluation of outcomes because both shirts and fish are
goods.
Diversification in the absence of exchange
If Alex cannot exchange the goods with others, does the best he can by following the mrs = mrt rule and finding the point on the production possibilities frontier that is tangent to the highest indifference curve, at point d (for
diversified production), and consuming xd and yd .
In this case Alex chooses to produce some of both goods because the
marginal utility of each of them is diminishing the more he consumes, so
having some of both is superior to having all of one kind. Remember this is
why his indifference curves are bowed inward toward the origin.
"The division of labor is limited by the extent of the market"
But, if the producer can exchange the goods with others, then there is a second way that he can "transform shirts into fish." He does not do so by reallocating his time from shirt making to fishing. Instead he can spend all of his
time making shirts and then exchange some shirts for some fish if he can find
a willing buyer for his shirts.
Suppose such a trader is found, and she is willing to buy any amount of his
shirts at a given price ( p): in return for p shirts she is willing to provide 1 kg of
fish. This is the second way of transforming shirts into fish, and the marginal
rate of transformation is p: the quantity of shirts that one has to give us in
exchange for a kg of fish.
This opportunity for exchange alters the feasible set constraining what Alex
can do, as shown in Figure 6.5. The orange line with the y-intercept at ȳ
(which is the maximum amount of shirts Alex can produce) represents his new
feasible frontier with exchange. Its slope is -p, the (negative of the) opportunity cost of acquiring more fish in terms of the shirts foregone. This means
that marginal rate of transformation of shirts into kg of fish is just p: giving up
p shirts gets you 1 kg of fish. Here the process depicted by movements along
R E M I N D E R Remember that production
functions with economies of scale are convex
(output increases at an increasing rate
with input) and production functions with
diseconomies of scale are concave (output
increases at a decreasing rate with input).
304
MICROECONOMICS
- DRAFT
y = 10
y = 10 = ys
s
yd = 4.5
Quantity of shirts, y
Quantity of shirts, y
Feasible frontier when
specializing in shirts
mrs(x, y) = mrt(x, y)
d
e
ye = 4.9
yd = 4.5
mrs(x, y) = mrt(x, y)
d
uA3
uA2
Feasible
outputs
0
xd = 1.65
xs = 0
x=5
uA1
xd = 1.65
Kilograms of fish, x
xe = 3.4
x=5
6.67
Kilograms of fish, x
(a) Economies of scale without trade
(b) Economies of scale and trade
the price line shows exchange in varying amounts, not shifting Alex’s own
labor from making shirt to fishing.
Because at the price p, 1 kg of fish is worth the same as p shirts, then the
value of all of the combinations of quantities of fish and shirts along the orange price line in Figure 6.5 b. have the same value. This is because the
value (expressed in number of shirts) of the fish purchased p · x must be equal
to the value of the shirts sold (which is just the number of shirts sold, ȳ
p · x = ȳ
uA2
Feasible
outputs
uA1
y) or:
y
(6.4)
The constrained optimization problem that Alex faces comes in two steps:
• Step 1: Decide on whether to specialize and if so, in which good; to do
this find the distribution of labor time between fishing and shirt making that
maximizes the value of one’s output then
• Step 2: Decide whether to exchange any of the goods produced, and if so
how many; to do this maximize utility subject the new feasible frontier given
by the goods produced and the relative price.
In Figure 6.5, Alex will decide to produce at point s (for specialized production)
at the intercept of his production possibilities frontier with the shirt axis to
maximize the value of his output at the relative price p. Then he will exchange
the shirts he produces for fish at the price p to reach point e (for exchange)
on the highest indifference curve in his new feasible set, uA
3 . This is where:
marginal rate of substitution = p = marginal rate of transformation by exchange
Figure 6.5: Production possibilities frontier
(PPF) with economies of scale. In Figure a, we
present the producer’s choice when they do not
have the opportunity to trade. Using the mrs = mrt
rule, he producer chooses the point at which their
indifference curve is tangent to their production
possibilities frontier at point d. In Figure b, we
show what happens if the producer can exchange
shirts for fish at some constant price ratio. In this
case, he can do better by specializing in shirt
production (good y) and then acquiring the fish
she desires through exchange, not by producing
them. The green shaded area is the enlargement
of his set of feasible levels of consumption of the
two goods. He produces shirts only at point s,
then exchanges the shirts on the market for fish,
taking him to his higher indifference curve, u3 at
point e. Notice that in selecting point s the producer
is not implementing the mrs = mrt rule. The
reason is that he does better at the corner solution,
producing none of the x good at all. But in choosing
point e by exchange, he is consuming both goods
(not a corner solution) and so the mrs = mrt rule
implements his constrained utility maximum.
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
Shirt production
Figure 6.6: Production possibilities frontier
(PPF) with diseconomies of scale. The fourquadrant graph shows a production possibilities
frontier with diseconomies of scale in production in
the top-right quadrant. The diseconomies of scale
depicted in the production possibilities frontier arise
from a relationship in the production technologies
from the two different sectors: fishing (bottom right
quadrant) and shirt-making (top-left quadrant). A
worker is constrained by how much of their labor
time they can dedicate to either producing fish or
producing shirts. Their labor constraint is depicted
in the bottom-left quadrant, shown as a limit of 10
hours of labor per day.
Shirts,y
10 hrs of labor
for shirts produces
8 shirts
Feasible frontier
5.66
1
y = 2.53(ls)2
Feasible
outputs
Labor for shirts, ls
10
5
5.66
Feasible
labor hours
ls = 5, lf = 5
305
Kgs of fish, x
Fish production
1
x = 2.53(lf)2
ls + lf ≤ 10
Constraint
on total 10
labor hours
Labor for fish, l
f
5
10 hrs of labor
for fishing produces
8 kgs of fish
Our example – Alex choosing what to produce – demonstrates two general
truths:
• If one or more production function with economies of scale is available, it
may make sense to specialize but
• this will be true only if there are others producing different goods and there
are opportunities for exchange, integrating specialized producers with
generalist users to coordinate the division of labor.
This is the basis of the interdependence of different producers within the
division of labor.
Checkpoint 6.2: The choice of what to specialize in
E X A M P L E In modern economies a household may specialize in providing labor with
some particular mix of skills, training and
experience to an employer.
• Redraw Figure 6.5 with a higher relative price of fish (so p, the number of
shirts that one must give up to get a kg of fish is now larger).
• Show that if p is sufficiently high, Alex will do better specializing in fish (indicate the amount he will produce, and the amount he will exchange).
Diseconomies of scale, diversification and exchange
In contrast with the production with economies of scale illustrated in Figure
6.4, Figure 6.6 illustrates the case of diseconomies of scale, in both fishing
and shirt making. The result is that Alex’s production possibilities frontier is
bowed outward from, or concave to, the origin. With diseconomies of scale,
as Alex shifts his labor input from producing shirts to producing fish, he gives
M - C H E C K The production possibilities
frontier with diseconomies of scale in both
production functions is concave toward the
origin, or bowed in.
306
MICROECONOMICS
- DRAFT
e
ye = 9
Quantity of shirts, y
y=8
mrs(x, y) = mrt(x, y)
Feasible
frontier
Price line
(feasible frontier)
yd = 5.7
d
s
ys = 3.6
uA3
uA2
uA1
xe = 4.5 xd = 5.7 xs = 7.2 x = 8
Kilograms of fish, x
up larger and larger amounts of shirts to get smaller and smaller amounts of
fish. This reflects the fact that with diseconomies of scale the marginal product
of his labor decreases the more labor he devotes to production of a good.
With diseconomies of scale the mrt (x, y) increases as labor is reallocated
from shirts to fish reflecting the idea of increasing opportunity costs.
If he did not have opportunities for trading goods, he would select point d in
Figure 6.7 with utility uA
2.
But he can do better if he decides what to produce knowing in advance that he
be able to exchange the goods he produces. So he will use the two-step constrained optimization procedure outlined above: first decide what to produce
so as to maximize the value of his output, then exchange goods to maximize
his utility.
But due to the diseconomies of scale the result is not complete specialization.
He will decide to diversify. He will produce at point s somewhere in middle of
his production possibilities frontier where his mrt (x, y) is equal to the relative
price p, putting labor into producing both goods, to maximize the value of his
output at the relative price p.
But the possibility of exchange expands his feasible set: the orange line is the
feasible frontier with production at point s and exchange possible at price p.
He will then exchange one or the other of the goods he produces for the other
– in this case exchanging fish for shirts – to reach the highest indifference
curve that is in his feasible set, uA
3.
Figure 6.7: The production possibilities frontier
for fish and shirts when there are diseconomies
of scale in production. If the producer cannot
exchange the goods with others, he does the
best he can by finding the point on the production
possibilities frontier that is tangent to the highest
indifference curve, at point d, and consuming xd
and yd . But, if the producer can exchange the
goods with others, the producer chooses the
production point with the highest value at that price,
and then exchanges output to maximize utility at
point s and then moves along the price line to point
e on u3 .
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
307
6.5 Comparative and absolute advantage
The question ”What should you specialize in?” seems to have an obvious
answer: ”Specialize in what you are best at.” The same would seem to go
for countries: they should specialize in what they are best at producing. But
what, exactly, does that mean? ”Better” than other people (o countries)? What
if you are not better than others at anything? Should you not specialize in
anything?
Differing opportunity costs and comparative advantage
Or, does “better” mean ”better than you are at other things that you could do”?
If that’s what it means – at least compared to how good others are in those
same things – then we are talking about comparative advantage.
To see what this means, suppose that a recent graduate, Brett, has started
a data science business. When Brett writes reports for his business, there
are two tasks: entering data in some digital format and generating graphs to
some detailed specifications using the digitized data. Lets say that each graph
requires 1,250 keystrokes of data.
Brett has the option of doing both tasks, or doing either one of them himself
and getting the other done for pay on Mechanical Turk (which calls itself
”the online marketplace for work”). There are many people like Brett, some
of them offering their services on M-turk, as it is called, and with their pay
purchasing other services from M-turk. But we will consider just one of these
people named April (there are lots of people like her too).
When we say April, we really mean "people like April ready to sell graphs
on M-turk, in return for data entry sold by people like Brett." And a similar
statement goes for Brett. The reason is that if we had some particular Brett
exchanging with a particular April, then in agreeing on a price they would
have to agree on the amounts to be exchanged (as Ayanda and Biko did in
Chapter 4). This would be an unnecessary complication, so we avoid it by
assuming that both Brett and April can purchase and sell as much as they like
at whatever price is posted on M-turk.
M-Note 6.2: Opportunity costs, feasible frontiers and comparative advantage
To understand how production functions, opportunity costs, and the feasible frontier
determine absolute and comparative advantage we use the following notation:
1
• ax
= time (fraction of an hour) required to input 1000 keystrokes of data ( 11
for April
1
and 10
for Brett, so April has an absolute advantage in data entry.)
1
• ay
= time (fraction of an hour) required to produce one graph ( 20
for April and 18 for
Brett, so April has an absolute advantage in graph-making)
O NLINE M ARKETPLACESMechanical Turk
is one of many online marketplaces for
work tasks for pay. Others would include
Clickworker, Fiverr, UpWork and many
others. People can be paid be for small,
short tasks like data input (which is more
typical for sites like M-Turk and Clickworker)
or for more advanced jobs like Fiverr and
UpWork.
308
MICROECONOMICS
Graphs made, y
20
- DRAFT
Figure 6.8: Feasible frontiers: absolute and
comparative advantage. April has an absolute
advantage in the production of both goods because
her feasible frontier is outside Brett’s. Brett has
comparative advantage in data entry because
his feasible frontier is flatter than hers (lower
opportunity cost of data entry). Without the
possibility of exchange Bret completes 4 graphs
(at point g). Remember: each graph requires
1250 keystrokes of data (also remember that
the horizontal axis of Figure 6.8 is measured in
thousands of keystrokes). This means that they
must be on the dashed orange line from the origin.
The question is how far out they can get. This
x
dashed line for y = 1.25
lets us see how many
complete graphs (data entry and graph preparation
combined) each person could be by themselves in
one hour.
ffA
Each graph needs
1,250 keystrokes
x
y=
1.25
8
ffB
i
6.11
g
4
5
10 11
7.64
Data entered ('000's), x
• x = thousands of keystrokes of data entered
• y = number of graphs made
• T = total time = 1 hour
Total labor time is composed of time spent entering data plus time spent producing
graphs, so the feasible set is defined by:
Time constraint
ax x + ay y  T
(6.5)
The equation for the feasible frontier is Equation 6.5 expressed as an equality and rearranged with y as a function of x. That is, the feasible frontier could be depicted as:
Feasible frontier
y
=
T
ay
ax
x
ay
(6.6)
Which means that:
dy
ax
=
= mrt
dx
ay
(6.7)
dy
dx is the negative of the slope of the feasible frontier (which can be seen from Equation
6.6). This is also the opportunity cost to either person for producing data entry in terms of
graphs.
But we can also use the numbers to find their opportunity costs. For April:
mrt (x, y) =
dy
ax
=
=
dx
ay
mrt (x, y) =
dy
ax
=
=
dx
ay
1
11
1
20
=
20
= 1.82
11
(6.8)
=
8
= 0.8
10
(6.9)
For Brett:
1
10
1
8
Because 0.8 < 1.82, Brett’s comparative advantage is in data entry.
Figure 6.8 shows how good April and Brett are at the two tasks, as indicated
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
309
by their feasible frontiers if they are restricted to working for just one hour. The
points making up their respective frontiers are the combinations of outputs
from data entry and graph making that use up one hour of their working time.
For example, in an hour Brett can produce 8 graphs and enter no data, or
10 (thousand) keystrokes of data entry, and no graphs, or 4 graphs and 5
(thousand) keystrokes, and so on.
Figure 6.8 illustrates the concepts of both absolute and comparative advantage. April has an absolute advantage in producing both goods: data entry
in thousands (x) and graphs (y). A person has an absolute advantage in the
production of a particular good if, given the set of available inputs, she can
produce more of it than some other person. M-note 6.2 shows how the feasible frontier is derived from the data on the productivity of April and Brett at the
two tasks.
In our case, absolute advantage means that, in one hour, if April devoted all of
her time to data entry, she could enter more thousands of keystrokes of data
than Brett (11 rather than 10) and likewise for making graphs (20 rather than
8). As her feasible set includes Brett’s entire feasible set (her feasible frontier
is farther from the origin), she can produce more in an hour than Brett can in
any combination – complete specialization in one or the other or some ratio of
data entry to figure making.
Different opportunity costs: The basis of specialization and exchange
This raises the question: if April is better at both data entry and graph making,
why would she want to trade with Brett at all? This is where the concept of
comparative advantage comes in. A person has a comparative advantage in
the production of a particular good if their opportunity cost of producing that
A BSOLUTE ADVANTAGE A person has an
absolute advantage in the production of a
particular good if, given the set of available
inputs, she can produce more of it than
another person.
C OMPARATIVE ADVANTAGE A person has a
comparative advantage in the production of
a particular good if the opportunity cost to
them of producing that is lower than it is for
another person.
good is lower than it is for another person.
For Brett, spending the hour it would require to enter 10 thousand more
keystrokes of data would mean that he could not make 8 graphs. So 8 graphs
is his opportunity cost of 10 thousand keystrokes of data entry. Translating this
to be in the units of the figure, 0.8 graphs is the opportunity cost to Brett of 1
thousand keystrokes of data entry (which is the negative of the slope of his
Maximum possible data entry
Brett
April
10
11
8
20
0.8
1.82
1.25
0.55
(thousands of keystrokes per hr)
Maximum possible graphs
(graphs per hr)
Opportunity cost of 1 thousand keystrokes
of data entry (in graphs)
Opportunity cost of making 1 graph
(in thousands of keystrokes)
Table 6.1: Absolute and comparative advantage:
Number of bits of data and graphs created in
one hours work. The entries in blue show that
April has the absolute advantage in producing both
data entry and making graphs. The entries in red
show that Brett has a comparative advantage in
producing data entry (0.8 < 1.82) and similarly April
has a comparative advantage in making graphs
(0.55 < 1.25). Remember, if they are working alone,
they would never produce only graphs or only data,
because they need a combination of graphs and
data for the project. Additionally, note that, for both
of them the two opportunity costs are simply the
1
inverse of one another, e.g. for Brett 1.25 = 0.8
.
310
MICROECONOMICS
- DRAFT
feasible frontier).
By contrast, for April, 10 thousand keystrokes of data entry requires just 55
minutes (she enters 11 thousand keystrokes per hour) and in that period of
time she could have made 18.2 graphs. So for April the opportunity cost of 10
thousand keystrokes is 18.2 graphs, or translating this to the quantities in the
figure, the opportunity cost of a thousands keystrokes is 1.82 graphs.
Brett’s comparative advantage is in data entry. This is not because he is so
good at data entry; April is better at data entry then him. It is because he is
so unproductive in producing graphs, so the opportunity cost of taking time
away from graph-making to do data entry (the graphs he otherwise could have
made) is low. It can similarly be seen that April’s comparative advantage is in
producing graphs.
Table 6.1 summarizes Brett and April’s absolute and comparative advantage
in these tasks.
Here is a simple way to remember the difference between absolute and comparative advantage:
• If for a given axis (horizontal or vertical) the intercept of one person’s
feasible frontier is outside (farther from the origin than) the other’s, then
that person has an absolute advantage in the good on that axis.
• Comparative advantage is determined by the slope of the feasible frontier :
The comparative advantage of person with the flatter feasible frontier is in
the good on the horizontal axis. This is because the (negative of the) slope
of the feasible frontier (how steep it is) is the opportunity cost of the good
on the x axis. The comparative advantage of the person with the steeper
feasible frontier is in the good on the vertical axis.
The second bullet says that unless the two feasible frontiers have the same
slope, the comparative advantage of the two people will differ. Even though
one of them may not have an absolute advantage in either good (like Brett)
each will have a comparative advantage in one of the goods.
So Brett has to be comparatively good at something. The data show that the
opportunity cost of data entry is less for Brett than it is for April. We now show
why this provides the basis for Brett specializing in data entry, April spending
all her time making graphs, and the two entering into an exchange.
Checkpoint 6.3: Comparative and absolute advantage
a. If April could make only 7 graphs in an hour (Brett’s productivity remaining
unchanged) which of them would have an absolute advantage in which of
the goods? In which good would April’s comparative advantage be?
b. If in an hour Brett could enter only 4,400 keystrokes of data, who would have
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
comparative advantage in data entry?
6.6 Specialization according to comparative advantage
In Figure 6.8 we can see that if Brett produced both data entry and graphs
himself, he would get to point g – four completed graphs based on the required 5 thousand keystrokes of data. Similarly, April working by herself could
produce 6.11 graphs along with the necessary data.
Will the two be able to do better by specializing in the production of just a
single good each, and then exchanging graphs for data entry, so that each
would have the required keystrokes of data for each graph?
The opportunity to exchange expands the feasible set
To see that they will, think about two ways that April can get data entered:
she can do this herself, with every thousand keystrokes entered bearing an
opportunity cost of 1.82 graphs not made. Or she could pay someone else to
enter data, paying with some of her graphs. What is the most she would be
willing to pay for a thousand keystrokes of data entered? The answer is 1.82
graphs which is what she would have to "pay" in graph-making foregone, if
she did the data entry herself. This is her maximum willingness to pay for data
entry. This is also the slope of her feasible frontier.
Would Brett be willing to sell her data entry for a price less than 1.82 graphs
per thousand key strokes? The lowest price at which he would sell a thousand
keystrokes of data entry is 0.8 graphs because this is his opportunity cost of
data entry. His opportunity cost is the number of graphs he gives up producing
if he enters a thousand more key strokes. This price is called Brett’s minimum
willingness to accept (giving up data in return for graphs). This is the slope of
his feasible frontier.
Because April’s willingness to pay is greater than Brett’s minimum willingness
to accept – the lowest price at which he would sell keystrokes – each of them
can benefit by specializing and then entering into an exchange.
In Figure 6.9 we show what Brett can accomplish when he specializes in
data entry and then exchanges some data entry for the graphs he needs to
complete his project. The exchange opportunities are shown by the price
lines (which are parallel because both people face the same relative prices).
The (negative of the) slope of the price line is the number of graphs that
can be purchased with a thousand keystrokes of data entry, or 1.45 in our
example illustrated in the figure. Steeper is better for Brett, flatter is better for
April.
311
312
MICROECONOMICS
20
- DRAFT
Figure 6.9: Feasible frontiers and relative prices
for exchange. Without the possibility of exchange
Brett completes 4 graphs (at point g). The two
arrows show that instead, he could move to point
sB , specializing entirely in data entry, and then
exchange some of his data entry with April in return
for her making 5.16 graphs for him. The arrows
at the top show how April could specialize and
exchange.
sA
Graphs made, y
14.5
Each graph needs
1,250 keystrokes
x
y=
1.25
price
line
8
7.11
j
i
ffB
5.16
h
g
ffA
sB
6.45
8.88 10 11
price
line
13.8
Data entered ('000's), x
Specialization and mutually beneficial exchange
To see why specialization and exchange will be mutually beneficial, you can
think of the following:
• The (negative of the) slope of the price line as as the marginal rate of
transformation of keystrokes into graphs by means of exchange.
• The (negative of the) slope of the feasible frontier is the marginal rate of
transformation of of keystrokes into graphs by means of devoting more time
to graph-making and less to data entry.
The possibility of exchange gives Brett a new feasible set, with the frontier
being the price line passing through any point on his "working alone" feasible
frontier, indicating his exchange opportunities when he can buy 1.45 graphs
with a thousand key strokes. With this new opportunity he could move in two
steps from point g to point h. He could do this if he, first, specialized at point
sB and then, second, engaged in exchange, moving up the price line to point
h.
Similarly, if April were at point i producing both goods, she could move to point
j if she specialized at point sA and engaged in exchange to take her down her
price line to point j. In Table ?? we compare their situation when producing
both goods with the outcome when they specialize and trade.
The reason why a mutually beneficial exchange would be possible is that the
price at which the exchange took place (1.45 graphs per 1000 keystrokes)
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
Brett
April
Working independently
Data entry
10,000 (1 hour)
15,280 (1.39 hours)
Making graphs
8 (1 hour)
12.22 (0.61 hours)
Graphs submitted
8
12.22
for the project
Specializing and trading
Data entry
20000 (2 hours)
0
Making graphs
0
40 (2 hours)
Work produced
12,900 keystrokes
14.22 graphs
7,100 keystrokes
25.78 graphs
Others’ work purchased
10.32 graphs
17,775 keystrokes
Project submitted
10.32 graphs
14.22 graphs
for own project
Work for pay to
exchange with others
was greater than Brett’s opportunity cost of keystrokes and less than April’s
opportunity cost of keystrokes. Or returning to Figure 6.8 the slope of the price
line was greater than the slope of Brett’s feasible frontier and less than the
slope of April’s.
We have not explained why this particular price was the one at which they
traded (as this would have been distraction from introducing comparative
advantage). But any price between 0.8 and 1.82 would have allowed mutually
beneficial exchange to take place.
What made specialization possible in this case is two things:
• Differences: Brett and April differed in their comparative advantage so
there was some price – 1.45 per 1000 is just one example – at which they
could both benefit form an exchange.
• Opportunities for exchange: There was a way to exchange one’s completed
tasks with others so as to obtain the right mix of data entry and graph
making.
Checkpoint 6.4: The distribution of the gains from specialization and
exchange
a. Using Figure 6.8, determine the price (graphs per thousand keystrokes) such
that Brett would not benefit at all from specializing and trading and also the
price ratio such that April would not benefit.
b. Use the required number of thousands of keystrokes per graph (1.25) to say
how many graphs Brett could make if the price at which he could sell 1000
keystrokes fell from 1.45 graphs to 1 graph.
313
Table 6.2: Specialization and exchange according to comparative advantage. The price of a
thousand keystrokes of data entry is 1.45 graphs.
Remember Brett is exchanging his data entry for
graphs with people like April (not just April herself).
And the same goes for April’s exchanges. This is
why it is possible for the number of graphs that our
particular Brett purchased (5.16) to differ from the
number of graphs that April sold (12.89).
314
MICROECONOMICS
- DRAFT
6.7 History, specialization, and coordination failures
Brett and April simply decided to specialize in the tasks in which each had
a comparative advantage. The existence of the online marketplace for tasks
made this possible, and both people benefited by comparison to producing
their reports without specializing. In the earlier example, Alex simply chose
to produce shirts rather than fish, and he was able to feed himself because
he could exchange shirts for fish. These personal examples have important
lessons about specialization.
But comparative advantage is more often applied to what countries do, not to
what people do. When it comes to countries, we cannot say that, for example,
Germany "decided" to specialize in machine tools and Bangladesh in textiles.
What countries specialize in is the result of decisions made by vast numbers
of people independently choosing what kinds of skills they will learn, the jobs
they will take, what kinds of products the firms they own will produce and
similar decisions. Countries – unlike Brett and April – can sometimes end
up specializing in such a way that they remain poor. Had they specialized in
something else, they would have been rich.
To see how countries can specialize and stay poor, return to the feasible
frontier in Figure 6.5. But now think about the figure as applying to an entire
country, not just choices that Alex might make between fishing and making
clothing.
In this case, the economies of scale in the production of shirts occur because
in every firm labor is more productive in producing shirts the more shirts
are being produced in all of the clothing industry. Industry-wide (rather than
firm-level) economies of scale are called economies of agglomeration.
Economies of agglomeration means that the productivity of labor is greater
the larger is the total output of the many firms producing similar goods in one
country or region.
Economies of agglomeration contribute to the geographical concentration of
particular industries, for example:
• software engineering in Bangalore (Bengaluru), India
• finance in Hong Kong, London and New York City
• information technology and IT related production in Silicon Valley, California
• machine tools and motor vehicles in the Stuttgart-Munich region of Germany
Economies of agglomeration occur because when large numbers of people
are employed in producing the same product, the skills and other knowledge
particular to that industry are widely diffused in the population, resulting in
E CONOMIES OF AGGLOMERATION refers to
cases in which the productivity of labor is
greater, the larger is the total output of the
many firms producing similar goods in one
country or region.
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
Quantity of shirts, y
yS
Figure 6.10: Production possibilities frontier
poverty trap. In the figure, the country is specialized in fish and point f and producing x̄F and
obtaining welfare (utility) of uF at point b, but would
like to produce shirts (which would take them to a
higher indifference curve). The country is better
off specializing in fish than they would be if they
produced a mix of goods, but they would prefer to
be producing ȳS shirts at point s where they would
be on the higher price line and therefore obtaining
higher utility at uS at point a (uS > uF2 > uF1 ).
The country cannot simply shift inputs to produce
new outputs as that would require a dramatic
re-purposing of production.
s
Price line when
specializing in shirts
Feasible frontier
(production possibilities frontier)
a
c
b
uS
Price line when
specializing in fish
uF2
uF1
Feasible outputs
f
xF
Kilograms of fish, x
higher levels of productivity across the board. Public policies favoring a locally
dominant industry also reduce costs.
If the relevant economies of scale do not pertain to an individual firm, but
instead are economies of agglomeration, then a single firm even if capable of
operating at a large scale will have little reason to, for example, introduce a
truck manufacturing plant in a finance agglomeration such as Hong Kong, or
an IT region such as the Silicon Valley.
In Figure 6.10, a country could find itself at point c on indifference curve uF
1
where they have diversified production or at b where they have specialized
production in fish, trade some of the fish along the price line at price p and
arrive at bundle b on indifference curve uF
2 . They are better off specializing
in fish than they would be if they tried to produce a diversified set of goods
as shown by the specialized production resulting in higher utility at b than at
c.
In contrast, if the very same country specialized in producing shirts and then
traded some of their output to acquire fish, they would consume at bundle a
on indifference curve uS . Suppose that for as long as anyone can remember
they have specialized in fishing. Why don’t they just change that? In the case
of April deciding to specialize in data entry, or Brett in making graphs, the two
people would have quickly realized that they were ignoring their comparative
advantage: they would quickly switch their specialization to what in which they
have a comparative advantage. But in the case of an entire country how would
they switch?
315
316
MICROECONOMICS
- DRAFT
Economies of scale and poverty traps as an assurance game
Player B
Suppose that in the fish-producing country a few people realized that every-
fish. You can see this because getting more shirts by producing them – that is
moving along the feasible frontier away from point x̄F – rather than producing
fish and trading them for shirts is a losing proposition. With specialization,
people would produce at x̄ and trade to point b on indifference curve uF
2 But,
if they clothed themselves by producing shirts rather than exchanging fish for
shirts the best they could do would be point c (diversified production) on a
lower indifference curve (uF
1 ).
A country specializing in fish in this model is locked in to lower income. If
they all decided to switch then they would all be better off. But as long as
the decision about what each person will produce is taken independently,
people would not specialize in shirt production. They are facing a coordination
problem similar to those discussed in Chapters 1, 4, and 5.
To see this, imagine that the population of the country we have been modelling is composed of just two people Anjali and Budi. Budi’s parents have
urged him to take up fishing, and Anjali’s parents, too, have urged her to continue with the family’s traditional livelihood.
To determine if each will take up fishing or shirt making they will engage in
the non-cooperative game shown in Figure 6.11. Assuming that each spend
5 hours a day working, we have calculated their output depending on their
choice and the choice of the other, using the production functions in Figure
6.4. So:
• Fishing for 5 hours will produce 2.5 kg independently of what the other
does, and the price of fish is 1, so the value of their output if either of them
fish is just 2.5.
• If both produce shirts, that is 10 total hours of shirt production resulting in
10 shirts, or 5 for each of them; at the price 0.67, they both receive a value
of output of 3.33 (5 ⇥ 0.67)
• If one produces shirts and the other does not, the production function for
1 s 2
shirts tells us that the output will be 10
(l ) , with l s = 5, this results in 2.5
shirts, with a value of 1.67 (2.5 ⇥ 0.67).
You can use the circle and dot method (introduced in Chapter 1) to identify the
Nash equilibria of the game. There are two: both fish or both produce shirts,
and producing shirts Pareto dominates fishing.
How would the two play the game? That would depend on their beliefs. Budi
Shirts
produce shirts they would be much worse off than specializing in producing
Player A
specializing in shirts. What could they do? If people decided individually to
3.33
Fish
Shirts
one would be better off (be on a higher indifference curve) if they switched to
2.5
Fish
3.33
2.5
1.67
1.67
2.5
2.5
Figure 6.11: Two people in a country have to
choose whether they will play "Shirts" or "Fish". At
given prices of p = $1 for fish and p = $0.67 for a
shirt, they confront the following payoffs. The game
is an assurance game with two Nash equilibria
(Shirts, Shirts) and (Fish, Fish) where (Shirts,
Shirts) is Pareto superior to (Fish, Fish) and Pareto
efficient.
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
317
might reason that for him taking up shirt making is risky because if Anjali does
not make the same choice then there will be no economies of agglomeration
and the payoff would 1.67. Fishing by contrast is a sure thing: 2.5. Anjali
might well think the same way. Based on the traditions of their society they
would probably believe that the other would take up fishing. And so they would
both fish.
Of course if they could have agreed to both produce shirts, then they would
have benefited from the economies of agglomeration and each produced
twice as many shirts (5) as one of them working singly could do. But we
are letting Anjali and Budi represent an entire population who are mostly
strangers to one another, not two neighbors who could agree on a course of
action.
So they have no way of coordinating their actions. Like the farmers of Palanpur – all planting late when they could all be better off by planting early – they
will be less well off because of the poverty trap which they cannot escape
because they lack institutions that would coordinate a joint decision.
This kind of self-perpetuating specialization is part of the reason why so much
of the world remained poor while other parts became wealthier. The labor
force of Africa, Asia and Latin America engaged in agriculture and other low
productivity sectors. Europe and its offshoots (North America, Australia, and
New Zealand) became wealthier starting in the early 19th century in some
measure by producing shirts and other manufactured goods.
In the late 19th century and into the 20th century other countries shifted their
specialization to sectors with higher labor productivity. This began with Japan,
and continued with South Korea, Singapore, China, and Vietnam. These
countries shifting to manufacturing as a higher labor productivity activity
mirrors our example of switching to shirts from agriculture. The modern manufacturing in these countries includes electronics, ship-building, and automobile
production.
In all of these cases the change in specialization occurred as a deliberate
government project, not as the result of countless people deciding to produce
commodities like shirts rather than fish.
6.8 Application: The limits of specialization and comparative advantage
Economies of scale and opportunities to exchange are pervasive in modern
capitalist societies, and, as a result, we live with an extensive (even global)
division of labor in which many individual households and firms specialize in
producing only one or a narrow range of products and meet their needs by
exchanging these products through monetary transactions.
Figure 6.12: A street-side dosa. Courtesy Sachin
Gupta. CC ShareAlike.
318
MICROECONOMICS
- DRAFT
When we think of specialization, we often conjure images of Silicon Valley’s
engineering and technology hub or the City of London financial center. But, India is home to one of the most developed and specialized information technology industries in the world based in Bangalore. The Bangalore based IT firms
InfoSys and Wipro exemplify the dynamics of an industry that grew from nothing in the early 1980s to become major global players by the early 2000s.2
Specialization occurs, too, in older industries. Manufacturers in Bangladesh
export a lot of shirts and hats, and very few bed sheets, whereas firms in
Pakistan export a great number of bedsheets, but very few hats.3 Neither is a
particularly skill-intensive kind of production and there is no reason for us to
expect that one of them ought to be better at bedsheets than hats. But they
have specialized due to the advantages of learning by doing and economies of
scale.
In contrast with this specialization, however, many households do still remain
diversified rather than specialized. Many households cannot achieve the
benefits of economies of scale due to insufficient wealth to sustain the training
and investment required for specialization, and also because of the riskiness
of starting businesses or engaging in just a single kind of work. As a result,
many poor households diversify of rather than specialize.
For example, Abhijit Banerjee and Esther Duflo describe the economic lives of
poor women in Guntur, a city in India.4 The women spend time in the morning
selling dosas (a rice and bean breakfast food), they make small amounts of
money collecting trash, they gather firewood to sell, they sell fruit, vegetables
and clothing (mostly saris), they make and sell pickles, or they work as shortterm laborers. Similar patterns of diverse occupations occur in Cote d’Ivoire,
Guatemala, Indonesia, Pakistan, Nicaragua, Panama, Timor Leste, and Mexico. An example from India shows one extreme: a survey by Nirmala Banerjee
in West Bengal showed that the average family had three people who worked,
sharing seven occupations among them.5
The economic analysis of these two different configurations of production –
specialization or diversification – is based on the same fundamental concept –
doing the best you can given a set of constraints. But as the examples above
show, whether a person or family specializes or diversifies is not simply a
matter of technology – economies or diseconomies of scale for example, or
learning by doing or differential skills. For a family with limited or no wealth
and exposed to uncertainty of their incomes in any single pursuit, risk mitigation becomes an important priority. As a result, diversification may be the best
they can do. We show in Chapter 13 how this very common combination of
limited wealth and exposure to uncertainty may contribute to the perpetuation
of poverty.
E X A M P L E Abhijit Banerjee and Esther Dulfo
won the Nobel Prize in Economics in 2019
(alongside Michael Kremer). They won the
prize for their work on projects to alleviate
poverty using the methods of randomized
controlled trials, which they have advocated
for worldwide to understand the impacts of
policy.
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
6.9 Production technologies
In modern economies, production takes place in families, in governments, in
privately owned firms and in other settings, each distinguished by a character-
319
P RODUCTION Production is the process by
which we transform the resources of the
natural world using already produced tools,
facilities, and inputs to meet human needs.
istic set of rules of the game determining who owns the goods produced, who
directs the production process and so on.
Private owners of the buildings, machinery, intellectual property and other
assets making up a firm aim to sell the output (which they also own) for more
than their inputs cost, the difference between sales revenues and costs being
the owners’ profit. The owners (or managers) of the firm choose the methods of production, the amounts of inputs it hires (hours of labor, number of
machines), and the level of output to maximize their profits, given the production methods available, the prices they pay for inputs, and the market prices
for their output. For this reason owners of firms want to minimize the costs
that they incur to produce any given level of output that they decide to produce. Here we explain how they choose cost-minimizing technologies to use
in converting raw materials and some given level of output of products for
sale. In Chapters 8 and 9 we turn to the owners’ decision about how much to
Inputs and outputs
Consider a firm producing an output, for example, cars, smartphones, or
clothing.
To produce its output, x, the firm needs to hire labor with the skills necessary
for the production tasks and provide the workers with raw materials, tools and
facilities. In a general model we could think of the inputs as a list describing
the amounts of labor of each kind and of all the different raw materials (wood,
steel, plastic, glass), tools (dies, drill presses, forges), and facilities (factories,
vehicles) required to produce the output. These inputs to the production
process are sometimes termed factors of production. We would measure all
these inputs over the same time period as output: so many hours of each kind
of employee per month, so much steel per month, so much factory space per
month, and so on.
It’s easy to see how the process works if we look at two dimensions on horizontal and vertical axes. The labor hired is l (on the horizontal axis), and k (on
the vertical axis) is the quantity of capital goods that the firm uses – the machines, tools and facilities that the firm needs to hire or own to produce their
output over the relevant time period.
We can describe one way of producing
a particular level of output by indicating in this space a level of the labor input
l , and the capital goods input k that will produce the specified output. This
combination (x, l, k ) describes one of the possible firm’s technique of production. For a given level of x, we can describe the technique of production
Quantity of capital goods, k
produce.
k1
A production
technique, (x, l, k)
i
l1
Hours of labor, l
Figure 6.13: A production technique, (x, l, k )
producing some amount of output x using labor
l1 and capital goods k1 .
FACTOR OF PRODUCTION Any input into
a production process is called a factor of
production. In the past economists often
referred to land, labor and capital goods as
primary factors of production, but this usage
is outdated given the essential role today of
other production inputs such as our natural
environment beyond "land" and knowledge.
T ECHNIQUE OF PRODUCTION A technique of
production is a particular way of producing
some given amount of output (x). In this
case it is a combination of an output level,
hours of labor input, and capital goods input,
(x, l, k).
T ECHNICAL EFFICIENCY A technique of
production is technically efficient if there
is no other technique with which the same
output can be produced with less of one
input and not more of any input.
320
MICROECONOMICS
- DRAFT
Quantity of capital goods, k
Feasible
k2
●
k1
●
f
i
●
h
Figure 6.14: Production techniques. For a
given level of output x = 100, we can describe
any technique of production as a point showing
the amount of labor, l , and the quantity of capital
goods, k, required to produce output x. The area
shaded in green shows the feasible combinations
of labor and capital goods that can produce output
x, which is equivalent to the feasible set introduced
in chapter 3. The area in blue shows the infeasible
combinations of capital goods and labor to obtain
an output of x = 100.
g
●
Infeasible
l1
l2
Hours of labor, l
as a point in (l, k ) space, as in Figure 6.13.
Technology and feasible production
The firm is constrained by the available technology, which describes what
techniques it can in fact carry out, given its state of knowledge, the skills of
workers, and the conditions of work (health, safety and intensity) that the firm
can legally and socially impose on its workers. Technology is therefore not just
a question of engineering or scientific knowledge, but also involves relations
between workers and management and among workers, and the legal and
institutional framework within which the firm operates.
Figure 6.14 displays the feasible set for producing a hundred units of output. The green shaded region shows the set combinations of capital goods
and labor, sufficient to produce a given output, x = 100, of the good x. The
dark green line is the border of the feasible set and is called an an isoquant.
There are additional isoquants each associated with the differing level of output that the inputs produce. So there is a set of isoquants, called an isoquant
map, each one of them derived from a production function and associated with
a different level of output.
Each of the four lettered points in Figure 6.14 is a particular combination of
labor and capital goods that are sufficient to produce 100 units of good x. But
the owners of a firm seeking to produce that amount would not be equally
happy to use any of the four.
I SOQUANT An isoquant gives the combinations of two inputs that are just sufficient
to produce a given level of output. ‘Same
quantity’ is exactly what the two parts of
the name isoquant mean: ‘iso’ for ‘same’
and ‘quant’ for quantity. The quantity that is
the same on the production isoquant is the
quantity of output.
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
321
k3
Steeper ray:
a●
more k−intensive
less l−intensive
k1
●
b
Flatter ray:
more
l−intensive
less
k−intensive
l1
xB = 100
Quantity of capital goods, k
Quantity of capital goods, k
xA = 100
k3
c combines
techniques
a and b
a●
k2
c
●
k1
l3
●
b
l1
l2
l3
Hours of labor, l
Hours of labor, l
(a) Comparing two techniques
(b) Combining two techniques
Technique i dominates technique h because it uses less of both inputs to
produce the same output. Similarly techniques f and g are dominated by
point i: each use the same amount as does technique i of one input and more
of the other to produce 100 units. Technique i is called technically efficient
because (considering the alternatives, f, g, and h) there is no other technique
that produces the required amount of output (x = 100) with less of one input
and not more of any other.
The isoquant map derived from a production function is analogous to the
indifference curves based on the utility functions in Chapter 3. It is important
to remember that an production isoquant is a constraint on the choice of
inputs required to produce a particular level of output, rather than something
to be maximized. A utility maximizer wants to get to the highest possible
indifference curve given the set of feasible options. The cost-minimizing firm
wants to get to the minimum cost point on the production isoquant for any
given level of output.
Figure 6.15 shows a production isoquant with two techniques of production
one of which uses more capital goods and less labor than the other. The
production isoquant includes the points representing the two techniques
and the line joining them, representing the possibility of doing some of the
production with one technique and some with the other. The second technique
of production provides some possibility of substitution of one input for the
other by switching from one technique to the other, but this substitution is
limited because there are only two techniques.
x = 100
Figure 6.15: Production isoquant combining two
techniques For a given level of output x, there may
be more than one feasible technique of production.
The production isoquant in this figure consists of
the two techniques, and the straight line between
them, representing production with a combination
of the two techniques (as shown by point c). The
availability of more than one technique implies that
substitution of one input for the other is possible
by shifting some production from a more capitalintensive technique to a more labor-intensive
technique.
R E M I N D E R A Pareto-efficient allocation is
one that is not dominated by any alternative, so there is no other allocation that is
preferred by at least one person and not
"dis-preferred" by any person. The definition
of technical efficiency is similar but applying
to techniques and inputs used rather than
allocation s of goods, and people’s utilities.
322
MICROECONOMICS
- DRAFT
As shown in Figure 6.14 and 6.15, the production isoquant can be thought
of as points corresponding to the various techniques of production, and the
lines connecting those points (which correspond to mixing the techniques of
production). The production isoquant is equivalent to the idea of the feasible
frontier in earlier chapters as it defines what combinations of labor and capital
goods can feasibly produce the level of output, x.
6.10
Production functions with more than one input
The techniques of production available are often described in a production
function, which is a mathematical expression giving the least quantity of inputs
– such as capital goods (k) and labor (l ) – that are sufficient to produce any
given level of output, x. The production function can also be thought of as
specifying the maximum level of output attainable for each combination of
inputs:
Production Function x
f (l, k)
=
(6.10)
You have already seen examples of the simplest production function in the
Leontief production function in which there is but a single technique available
for a given level of output x as in Figure 6.14 and Figure 6.16.
R E M I N D E R We have already examined
production functions, but so far they have
only involved one input, such as labor as an
input into studying in Chapter 3 or labor as
an input into either fishing or shirt production
in section 6.2.
P RODUCTION FUNCTION A production
function x = f (l, k ) describes a firm’s
available set of techniques of production as a
mathematical relationship. Here we present
production functions with just two inputs –
labor and capital goods – but production
functions may describe the relationship
between output and any number of inputs,
labor with different skills, for example, or
different kinds of capital goods (buildings,
machines, and so on).
As in those figures what are called Leontief production isoquants are rectangular because the technology specifies a given ratio of capital goods to labor
(at the point of the rectangular isoquant). If that particular ratio of inputs is in
use, then adding more labor or more capital goods has no effect on production: their marginal products are zero. There are therefore no possibilities of
substituting one factor of production for another.
To clarify what this "no substitution" assumption means with an extreme example, think about nuts and bolts: if you have n nuts and n bolts, then having
n + 1 bolts is no better than having n bolts. A bolt is useless without a nut, and
a nut is useless without a bolt. You need to use the inputs in fixed proportion
to each other to get “a nut and a bolt.”
M-Note 6.3: Leontief Production Function
The output of good x, is produced with l the amount of labor input used and k the amount
of capital goods used. al and ak are the minimum amounts of labor and capital goods
required to produce a single unit of output.
Noting that min(m, n) means m and/or n, whichever of m or n is least (or both of them if
they are equal), the Leontief production function can be written:
x = f (l, k) = min
✓
l k
,
al ak
◆
(6.11)
The equation can be read: "The number of units of x produced is the smaller ("min") of
the ratio of the amount of the input used (the numerator in the two fractions) to the input
required for a single unit of production (the denominator) Any capital goods input in excess
H I S TO RY Wassily Leontief (1906-1999) was
a Russian-American Nobel Laureate in economics. He modeled the whole economy as
what became known as an input-output system, with each industry being represented
by a Leontief production functions. His work
is valued by economists because it allowed
a mathematical representation of the whole
economy that could be estimated empirically
(for example, engineers could determine how
many tons of coal are needed to produce a
ton of steel.)
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
323
of the minimum amounts required is of no use in production, and might as well be thrown
away, and similarly for any labor input.
Checkpoint 6.5: Leontief Production
a. Using the Leontief production function in the M-Note, if al = 2 hours and
ak = 1 hour of machine use, what is the level of output in each of the following cases:
i. l = 10 and k = 10
E X A M P L E Leontief’s input-output models
are today used, for example, to calculate the
amount of CO2 emissions produced per unit
of output of each industry, taking account of
both the direct and the indirect inputs. That
is counting for example not only the coal
used to produce a ton of steel, but the coal
used in producing the machinery and all of
the other inputs required for a ton of steel.6
ii. l = 10 and k = 5
iii. l = 16 and k = 5
b. In each case above (i, ii, iii) how would output change if one more hour of
labor or one more unit of machine time were devoted to production (this is
the marginal product of labor and of machine time, respectively)?
Cobb-Douglas production function
Another representation of how inputs are combined to produce outputs is the
Cobb-Douglas production function.
Cobb-Douglas Production Function
x(l, k)
= ql a kb
R E M I N D E R The Cobb-Douglas production
function has the same structure as the
Cobb-Douglas utility functions we studied in
Chapter 3.
(6.12)
The Cobb-Douglas production function requires that l > 0, k > 0 for production
to take place: some of both inputs are essential, but their proportions used
can vary.
• 0 < a and 0 < b capture the contribution of labor and capital goods,
respectively to producing output;
• The sum of a and b tells us how output responds to changes in propor-
H I S TO RY Paul Douglas (1892-1976) developed the function with his colleague at
Amherst College, Charles Cobb. Though
a Quaker, Douglas was fiercely anti-fascist
and during World War II volunteered for the
U.S. Marine Corps as a private at the age
of 50. He later won two purple hearts in
recognition of the battle wounds he suffered
in the Pacific theatre. He went on to be a
prominent member of the Democratic Party
and a U.S. Senator serving from 1949-1967.
tional increases in both of the inputs indicating whether the firm experiences economies of scale, diseconomies of scale, or constant returns to
scale.
• q > 0 is a positive constant that captures a level of productivity of the
specific technology .
A Cobb-Douglas isoquant for x = 100 is shown in Figure 6.17. The negative
of the slope of an isoquant at any point is the ratio of the marginal products of
labor and capital goods inputs, and is called the marginal rate of technical
substitution, or mrts(l, k ).
The marginal rate of technical substitution
The negative of the slope of a production isoquant shows the ratio in which
the two inputs can be substituted for each other while output remains constant, the marginal rate of technical substitution between the inputs. The
M ARGINAL RATE OF TECHNICAL SUBSTI TUTION The marginal rate of technical
substitution is the rate at which labor and
capital goods inputs can be substituted holding constant firm output. It is the negative
of the slope of the production isoquant and
equal to the ratio of the marginal products of
the inputs.
324
MICROECONOMICS
- DRAFT
Figure 6.16: Production isoquant given a CobbDouglas production function. The feasible set
of production for a given x is the set of techniques
of production, combinations of labor input and
capital goods, (l, k ) that permit the firm to produce
x. Given the Cobb-Douglas production function
x = ql a kb we can isolate k to find an equation
10
Quantity of capital goods, k
9
8
x feasible
with (l, k)
7
for the production isoquant: k =
⇣
x
ql a
⌘1
b
production isoquant has a negative slope.
6
Cobb−Douglas
isoquant
5
4
x = f(l, k)
x infeasible
3
with (l, k)
2
1
0
0
1
2
3
4
5
6
7
8
9
10
Hours of labor, l
marginal rate of technical substitution based on the isoquant is analogous to
the marginal rate of substitution, the negative of the slope of an indifference
curve, which we introduced in Chapter 3. But notice that in the case of the
firm seeking to minimize the cost of producing a given level of output, the
isoquant is the co nstraint not the firm’s objective. It tells the owners of the
firm what combinations of inputs will produce the given level of output (the
constraint).
M-Note 6.4: The marginal rate of technical substitution and marginal
products
The production isoquant is defined as the combination of inputs that can produce a given
output, f (l, k ) = x.
To find the slope of an isoquant we proceed as we did when finding the slope of an indifference curve. We use the property of the isoquant that the points on it made up of
different amounts of l and k result in the same level of output x. So for small changes in l
and k the following is true:
dx = fl (l, k)dl + fk (l, k)dk = 0
(6.13)
Because along a production isoquant the difference in output is zero (just like along an
indifference curve in earlier chapters the difference in utility is zero), Equation 6.13 can be
understood as follows:
fl (l, k)dl + fk (l, k)dk = 0
| {z } | {z }
Change
in x as
l changes
which we can rearrange as
mrts(l, k) =
Change
in x as &
k changes
dk
f (l, k)
= l
dl
fk (l, k)
(6.14)
. The
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
Figure 6.17: A production isoquant for the
Cobb-Douglas production function f (l, k ) =
l 0.5 k0.5 for the output level x = 4. The marginal rate
of technical substitution is the ratio of the marginal
x
mp
products, mp = x l , which is the negative of the
10
Quantity of capital goods, k
9
8
mrts(l, k) =
●
a
k
8
=4
2
mrts(l, k) =
4
=1
4
5
4
mrts(l, k) =
●
b
2
= 0.25
8
3
2
●
c
Cobb−Douglas
isoquant x = f(l, k)
1
0
0
1
2
3
4
5
6
7
8
9
10
Hours of labor, l
Equation 6.14 can be stated as:
Marginal rate of technical substitution
=
Marginal product of labor
Marginal product of capital
This is the negative of the slope of the production isoquant.
Checkpoint 6.6: Isoquants and marginal rate of technical substitution
Using the same production function as Figure 6.17, calculate
a. the marginal product of labor and the marginal product of capital
b. determine the mrts(l, k )
c. choose 3 different points (not a, b and c) along the production isoquant at
which to evaluate the mrts(l, k ) to confirm that mrts(l, k ) decreases as l
increases.
M-Note 6.5: Cobb-Douglas economies of scale
We start with the following Cobb-Douglas production function:
x(l, k)
=
k
slope of the production isoquant, dk
dl . Three points
along the isoquant curve are shown: a, b, and
c illustrating how the marginal rate of technical
substitution decreases moving rightwards down
the production isoquant from l = 4, to l = 1 to
l = 14 . Three gray dashed lines are tangent to the
production isoquant, the slopes of which are the
marginal rate of technical substitution, mrts(l, k ), at
each point.
7
6
325
ql a kb
To confirm what a firm’s economies of scale are we need to increase both inputs by some
proportion, S. Therefore, increase l and k by the proportion S. That is, multiply each input
by S before raising the input to the relevant power:
326
MICROECONOMICS
- DRAFT
x(Sl, Sk)
=
q(Sl )a (Sk)b
=
qSa l a Sb kb
=
Sa +b ql a kb
=
Sa +b x(l, k)
Now, take each S out of the parentheses:
x(Sl, Sk)
The final step occurs because we know that ql a kb is equal to our original production
function, x. If a + b is greater than one, the output grows more than proportionally with an
increase of l and k by the proportion S. Therefore, the production function has increasing
returns to scale. If a + b is lower than one, the production function has decreasing returns
to scale. If a + b is equal to one, it has constant returns to scale.
Diminishing marginal products of inputs
It is important not to confuse economies and diseconomies of scale, which
describe what happens when all inputs are changed proportionally with diminishing or increasing marginal productivity of one input when the others
are held constant (say, increasing labor, holding capital goods inputs constant).
A production function may have diminishing marginal productivity to any
one input when the others are held constant, and still exhibit economies of
scale when all the inputs are changed together.
M-Note 6.6: Diminishing marginal productivity
To compute the marginal product of labor in the Cobb-Douglas production function we
start with the production function:
x(l, k)
=
ql a kb
To find the marginal product of labor, we calculate the first partial derivative of the production function with respect to labor, which gives us the effect on total output of a small
change in the labor input, holding constant the level of capital goods input:
∂ x(l, k)
= xl = MPl
∂l
=
=
=
aql a
1 b
k
aql a kb
l
ax(l, k)
l
For a > 0 and l > 0, the marginal product of labor is positive: as you can see from the
equation immediately above, it is equal to a itself times the average.
To work out whether the marginal product of labor is diminishing, we need to know
whether the derivative of the marginal product of labor with respect to the labor input itself
M - C H E C K For the Leontief production
function we cannot compute the marginal
rate of technical substitution from the slope
of a production isoquant, because it’s slope
is undefined at the kink in the isoquant.
But at the "kink" in the isoquant, adding
more capital goods or more labor has no
effect on output, so we could view the
Leontief production isoquant as representing
an extreme form of diminishing marginal
products.
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
is positive, zero, or negative:
∂ 2 x(l, k)
= xll
∂ l2
=
a (a
=
=
1)ql a
2 b
k
1)ql a kb
l2
a (a 1)x(l, k)
l2
a (a
The sign of xll depends on the size of a .
Diminishing If a < 1, xll < 0 because a
ductivity of labor.
Constant If a = 1, then a
tivity of labor.
1 < 0, which implies diminishing marginal pro-
1 = 0, and xll = 0, which implies constant marginal produc-
Increasing If a > 1, xll > 0 because a
tivity of labor
1 > 0, which imply increasing marginal produc-
Checkpoint 6.7: Marginal products of factor inputs
Check your understanding by doing the following:
a. Repeat the steps in M-Note 6.6 to find the marginal product of capital goods
with a Cobb-Douglas function.
b. With b = 0.4 is the marginal product of capital goods diminishing, constant
or increasing?
c. Determine the values of b under which the marginal product of capital goods
will be diminishing.
6.11
Cost-minimizing technologies
Having introduced a description of the production process – the production
function – we now introduce the firm as a profit-maximizing entity. To determine the level of output that will yield the greatest profit for the owners of the
firm, consider two pieces of information that the owners of the firm would
need:
• Cost minimization: for every possible level of output, given the costs of
using the inputs to the production function, find the technique of production
that minimizes the costs of production;
• Profit maximization: using the resulting cost curve (describing the least
cost at which each level of output can be produced) and the demand curve
for the firm’s product, determine the level of output to produce.
Here we describe cost minimization. We describe profit maximization step
in Chapters 8 and 9. We call any particular combination of labor and capital
goods used (l, k ) as a bundle of inputs. Finding the minimum cost bundle
for producing each level of output the firm’s owners might want to produce
requires three steps:
327
328
MICROECONOMICS
- DRAFT
Figure 6.18: Three isocost lines are presented:
c1 , c2 and c3 . Isocost curves closer to the origin
are made up of less costly input bundles. The
equation for an isocost curve is given by c =
pk k + wl , where pk is the cost per unit of renting
capital goods, k is capital goods input, w is the
wage, and l is the quantity of labor input. We can
re-arrange this equation⇣in terms
of the capital
⌘
Quantity of capital goods, k
c3
pk
c2
pk
Marginal rate of transformation:
w
mrt(l, k) =
pk
goods input, k =
c
pk
w
pk
l . The slope of
the isocost line is determined by the marginal
rate of transformation of capital goods into labor,
mrt (l, k) = pw = dk
dl , which is the opportunity cost
c1
pk
k
of using more labor in terms of the lesser quantity
of capital goods that can be used, in order to hold
constant the cost of the resulting bundle.
c1
c2
c1 w
c3
c2 w
c3 w
Hours of labor, l
• Step 1: Calculate the cost of every input bundle that the firm might use
• Step 2: Identify bundles that cost the same, and use the resulting isocost
line to distinguish between more costly and less costly bundles and
• Step 3: Use the isoquants based on the available production functions to
determine, for each level of output, the least costly bundle.
Isocosts: Equally costly bundles of inputs
We assume that:
• the capital goods used by the firm are rented (for example, buildings and
equipment) rather than owned; and
• the firm’s own demand for labor and capital goods does not influence the
price it pays for these inputs (as would be the case, if the firm is small
in relation to the markets for its inputs, labor and various types of capital
goods).
Then the cost using any particular combination of labor and capital goods
depends on:
• Wages (w) paid per hour for the for the hours of labor hired (l )
• The rental cost of the capital goods ( pk ) times for the quantity of capital
goods used (k).
Then the cost of a bundle of inputs is:
c(l, k)
= wl + pk k
(6.15)
Quantity of capital goods, k
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
a
ka
329
Figure 6.19: The minimum cost of producing a
given level of output. To produce the output given
by the isoquant x, the least cost input bundle is
indicated by point i.
mrts(l, k) = mrt(l, k)
mpl w
=
mpk pk
b
kb
d
kd
c1
la
isoquant
x=x
c3
c2
lb
ld
Hours of labor, l
Using Equation 8.2, we know the cost of every input bundle so we can
construct an isocost line, a line showing all the possible combinations of
amounts of labor and amounts of the capital good that result in a constant
or equal (“iso”) level of costs. Re-arranging equation 8.2, we can find the
equation for an isocost line:
Isocost line k
=
c
pk
✓
w
pk
◆
l
(6.16)
The isocost lines represent the objectives of the owners of the firm. The
owners would like to find the way of producing their product that (for some
given amount of output) will put them on the lowest isocost line, that is, the
one close to the origin. The constraint limiting the owners decision is the
available technology or technologies as described by the production isoquant
I SOCOST LINE The line through an input
bundle (l, k ) when the wage is w and the
price to hire capital goods is pk is the line
through the input point with slope equal
to w/pk , and represents all the input
combinations that have the same cost.
R E M I N D E R In earlier chapters the indifference curves bowed in towards the origin
(like the green isoquant in Figure 6.19)
represented the objectives of the person,
that is the thing that she wished to maximize
based on their preferences, subject to some
constraint, for example a limit on how much
she could spend. Here the blue isocost lines
represent the objective of the firm’s owners,
that is the thing they wish to minimize, while
the curved isoquant is the constraint based
on the feasible set of production techniques
that produce at least the given amount of
output.
for the given level of output.
The general principle of cost minimization
Contrast point b with points a and d. At a, the marginal rate of technical substitution is high (the isoquant is steep), meaning that it can reduce capital
goods inputs substantially and still sustain the same level of output with only
a modest addition of the labor input. At the going cost of renting capital goods
and the wage rate and the firm would decrease its costs if it employed fewer
capital goods and more labor. The effect of this is to lower the marginal product of labor and raise the marginal product of capital goods. It would continue
to substitute labor for capital goods until the point where the ratio of marginal
products equals the ratio of prices for the inputs at b.
The isocost line and production isoquant toolset allows us to understand cost
P RINCIPLE OF C OST M INIMIZATION A firm
with a production isoquant consisting of
a continuum of techniques of production
defined by a production function x = f (l, k )
will minimize its costs at the point where
its marginal rate of technical substitution of
f
capital goods for labor, (mrts(l, k ) = f l ),
k
equals its marginal rate of transformation,
or the price ratio of labor for capital goods,
mrt (l, k) = pw . The principle of cost mink
imization is satisfied where the production
isoquant is tangent to the lowest isocost line.
330
MICROECONOMICS
- DRAFT
minimization. The firm wants to choose the lowest possible isocost line to
produce the output necessary to produce the target output, the minimum
cost technique of production.
Minimizing the cost of producing some hypothetical level of output involves
the principle of constrained optimization from Chapter 3. The constraint is
the isoquant for this particular level of output and the objective is to reach the
lowest isocost line.
Parallel to the principle of demand in chapter 3 for people maximizing utility,
we have a principle of cost-minimization for firms choosing techniques of
production. The firm will produce where mrts(l, k ) = mrt (l, k ) or where the
f
ratio of marginal products equals the ratio of input prices, f l = pw .
k
k
The cost-minimization graph and the graph describing the utility-maximizing
choice of a consumer facing a budget constraint are similar. But it is important
to recognize that the meaning of the elements is different. The constraint in
the production case is that the firm produce some given amount x = x, not the
isocost line. The owners of the firm are trying to move as close to the origin as
possible within the constraint that it produces the specified amount, while the
consumer is trying to move to as high an indifference curve as possible within
the constraint of the budget available. The iso-cost lines are analogous to the
indifference curves of the consumer (costs are being minimized, like utility was
being maximized), and the production isoquant is analogous to the feasible
frontier (it is the constraint).
Checkpoint 6.8: Choices of capital goods and hours of labor
Make sure you understand Figure 6.19 by explaining why, if the firm were producing at point d, it could reduce costs of producing the given amount of output
by using more capital goods and less labor.
Input prices and the choice of a labor-intensive or a capital-intensive technology
Now, think about a firm that sells some product and is considering which of
two technologies to use producing it. One uses some powerful machinery
(the capital good) and little labor while the other technology uses lots of labor
and a smaller machine. For concreteness, think of the two technologies as
similar to plowing a field using a powerful tractor or with a small garden type
roto-tiller.
If these two alternative ways of producing the good were described by a
Leontief technology, then we could say that the one using the roto-tiller is the
more labor intensive, or what is the same thing (because there are just two
inputs) the less capital goods intensive. In the Leontief technology, the ratio
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
Isoquant F
xf(lf, kf) = x
ki
Isocost curves
(high wage)
cH
1
kf
Isoquant G
xg(lf, kf) = x
Quantity of capital goods, k
Quantity of capital goods, k
cH
2
f
Isocost curves
(low wage)
kg
i
mrtsB(l, k) > mrtsA(l, k)
xBl
cHʹ
1
xBk
kh
j
kj
xAk
lh
of output, that is the ratio of the amounts al /ak , is a measure of the labor
intensity of the technology. While the more accurate expression is to refer to
capital goods intensive technologies, to save words we sometimes refer to
technologies as "capital-intensive."
Figure 6.20 a. illustrates this case first with Leontief technologies and Figure
6.20 b illustrates Cobb-Douglas technologies. In Figure 6.20 b, the CobbDouglas technology indicated by point g is more labor-intensive than the
technology at point f.
Where substitution between inputs is possible – as with the Cobb-Douglas
technology the distinction between labor-intensive and capital-intensive techintensive technology is the one that the owners of a firm would choose to
would minimize costs if wages were low relative to the cost of capital goods.
A capital-intensive technology, analogously, is one that would be used by
a cost-minimizing firm if wages were high relative to the costs of capital
goods.
Point f in Figure 6.20 a. shows the inputs required to produce a single unit
of output using the capital-intensive technology. Point g shows the same
information for the labor-intensive technology. Which technology the firm
will adopt in order to produce its product at the lowest cost depends on the
relative cost of labor and capital goods, as indicated by the isocost lines in
li
lj
Hours of labor, l
(b) Two Cobb-Douglas technologies
of inputs of labor to the inputs of the capital good required to produce a unit
nology is not so simple. The basic idea, however, is the same: the labor-
Labor−intensive
technology B
xB⎛⎝l, k⎞⎠ = x
cLʹ
1
cL2
(a) Two Leontief technologies
green and blue.
xAl
Capital−intensive
technology A
xA⎛⎝l, k⎞⎠ = x
lg
Hours of labor, l
lf
>
h
g
cL1
331
Figure 6.20: Choosing a capital-intensive or
labor-intensive technology to minimize costs.
The cost-minimizing choice of technology depends
on the wage and the cost of capital goods. Higher
wages (a steeper blue isocost lines) will lead
the owner to implement the more capital goods
intensive technology. In panel b the coefficients
for the labor-intensive Cobb-Douglas technology B
are a = 2/3, b = 1/3, and for the more capital
goods-intensive Cobb-Douglas technology, A
a = 1/3, b = 2/3.
332
MICROECONOMICS
- DRAFT
If wages are low, then the isocost lines are flatter, as shown in the figure with
the green isocost lines. If the firm uses the labor-intensive technology it will
incur costs of cL1 which is less than the cost it would incur if it used the capitalintensive technology when there are low wages (along cL2 ).
Higher wages (for the same rental cost of the capital good) are indicated by
the steeper isocost lines in blue. Using the labor-intensive technology with
higher wages (along cK
2 ) would incur higher costs than using the capitalintensive technology (along cK
1 ).
Figure 6.20 b. shows an analogous situation with greater substitutability between the two factors of production with Cobb-Douglas technologies. Once
K0
again, the relative costs are shown by two iso-cost lines, cL0
1 and c1 . The
unit isoquants show the different combinations of capital goods and labor
that would produce the same output, x. The owners of the firm would choose
point h if wages were high and point j if wages were low. Along the ray going
through points h and i, the ratio of kl shows that Technology A is capital-
intensive. At points h and i, the technical rate of substitution differs between
the two isoquants (which we can see with the labor-intensive technology B
having a much steeper isoquant at point i than the capital-intensive technology has at point h).
Checkpoint 6.9: Capital-intensive and labor-intensive technologies
• In Figure 6.20 panel a, show that there is one ratio of wages to the cost of
capital goods such that the least cost of producing x will be the same using
the two technologies.
• Show that the input price ratio the of firm using technology F will use less
labor and more capital goods than the firm using technology G.
• Show that if a firm had just two technologies to choose from, the Leontief
technology F from panel a and the Cobb-Douglas technology A from panel
b, it would choose the Cobb-Douglas technology if wages were either very
high relative to the cost of capital or very low. But for some input price ratio in
between, it would choose the Leontief technology.
• Explain why this means that it is not always possible to designate a technology as more labor-intensive or more capital-intensive.
6.12
Technical change and innovation rents
Firms and their owners compete not just by adjusting output levels but also by
seeking to innovate, either by finding new lower-cost techniques of production,
or by creating new products that open up new industries and new sources of
demand.
Quantity of capital goods, k
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
kb
Figure 6.21: Isocosts and technological
progress. A firm innovates to reduce its costs.
The firm starts at a with its initial labor and capital
goods combination (la , ka ) on isocost c2 . With
innovation the firm’s constraint is eased resulting
in a new production isoquant with an expanded
feasible set of production. As a result, the firm can,
at going prices of labor and capital goods, w and
pk , employ a lower quantity of capital goods and
less labor to produce output x at a lower cost total
cost, moving to a lower isocost line. Following the
principle of cost minimization, the firm chooses
the point at which its new production isoquant
is tangent to the lowest possible isocost at b,
employing (lb , kb ).
a
ka
initial
isoquant
x=x
b
c1
lb
333
innovation
isoquant
x=x
c2
la
Hours of labor, l
Viable innovations
The theory of cost minimization suggests an important insight into the causes
of innovation in production. A new technique of production will be of interest
to owners only if it lowers costs of production given current input prices, the
wage, w and the rental price of capital goods, pk . The introduction of a new
machine or a new organization of the production process may improve on
some existing available technique of production but it will be irrelevant to the
firm unless it results in lower costs than the existing minimum cost technique
of production.
How would we represent technological progress with production isoquant
curves? With technological innovation the firm should be able to produce the
same amount of output at lower total costs. We present production isoquants
with technological innovation in Figure 6.21. The initial technology is shown
with the cost-minimizing point a where the firm employs the combination
of labor and capital goods (la , ka ). With technological progress, two things
occur:
• The firm is able to produce the same amount of output with fewer inputs.
Its feasible set enlarges, resulting in a new production isoquant closer to
the origin.
• As a result of the new production isoquant, the firm will minimize its costs
at existing input prices and move to a lower isocost c1 , finding the point of
tangency of the isocost and the new production isoquant at point b.
E X A M P L E Forbes magazine produces
a list of the most innovative firms in the
world (https://www.forbes.com/innovativecompanies/list/. In 2018 Netflix, Tesla
(electric vehicles), Facebook, and Amazon
were in the top ten as was Hindustan
Unilever (a consumer goods producer
and marketer in India), and Naver (selling
computer and web services based in South
Korea).
334
MICROECONOMICS
- DRAFT
Notice that the new technology has made both labor and capital goods more
productive.
Innovation rents
Other things equal – importantly the prices of the inputs and its output – if the
firm produces the same quantity of x with lesser amounts of inputs per unit
of x, it would necessarily increase its profit. The firm would therefore obtain
an innovation rent. This is a rent because the firm’s next best alternative –
its fallback position – would be to not innovate. The innovating firm can lower
its prices and capture a larger share of the market. Other firms will either
make losses and exit the industry or innovate as well. As other firms imitate
the innovating firm, they would experience lower costs and would compete
with the innovator resulting in a lower price for the good and therefore lower
economic profits for the firm that initially innovated.
Innovation may result in new technologies that are more capital-intensive or
more labor-intensive. If the production technology is Cobb-Douglas, then the
innovation could be represented not only by an increase in q but also by new
levels of a and b , which would indicate the change in the importance of the
corresponding input.
A firm that successfully innovates by lowering its cost of production raises its
maximum profit. The firm’s increase in profit is an economic rent, or an innovation rent. It is an innovation rent because the firm’s fallback is the profit it
would have made at the initial technology, such as that resulting from point a
in Figure 6.21. The innovating firm will continue to obtain higher profits until
its competitors adopt the same or equivalent cost-reducing technical improvements. Once firms producing identical or similar products have matched the
innovator’s lower costs, if competition among firms is sufficient, some of them
will reduce prices to gain a larger market share, forcing other firms to do the
same. This will reduce innovation rents.
6.13
Application: What does the model of innovation miss?
Our model of innovation captures essential parts of the process by which
technical change revolutionizes an economy. But as this example shows, it
misses important aspects too.
A cluster of small firms in Sialkot, Pakistan produce about forty percent of the
worlds soccer balls - 30 million of them per year - including the match balls
for the 2014 World Cup. The industry is highly competitive not only among
the hundred or so firms in Sialkot, but also on a world scale, with Chinese
firms recently challenging the Pakistani dominance in the field. Firm owners
are constantly on the lookout for ways to slash costs. As the artificial leather
that the balls are made from constitute almost half the cost of a soccer ball,
E X A M P L E Innovation rents play the key role
in determining the profitability and survival
of firms. Apple, for example, keeps ahead of
its competitors by being the first to introduce
important innovations like the iPad or the
iPhone X (with facial recognition). Business
history also provides dramatic examples,
such as IBM in the 1980s where a firm that
has managed to maintain innovation rents
for many product cycles loses its position by
misjudging the next turn of the technological
revolution.7
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
335
they are particularly on the lookout for waste-saving methods of cutting the
pentagons and hexagons that make up the balls.
An Italian architect and her husband, an American economist, discovered a
way to cut the pentagons and hexagons from the large sheets of leather that
would allow a considerable saving of leather. (Unwittingly they had "discovered" what is called a "packing" principle already known by mathematicians.)
They found a tool and die-maker in Sialkot to make some test dies (a cutting
tool) using the new technique, expecting that they would quickly be taken up
by the cost-conscious firms.
In May 2012 they gave 35 firms the new technology. They calculated that the
new technology would increase profits of the companies adopting it by 10
Figure 6.22: [
0cm]A soccer ball. Using the new technology
the white hexagons and black pentagons making
up the ball could be cut with less leather wasted.
percent. Fifteen months later only 5 of the firms had made any substantial use
of the new cutting dies.8 While the new design was easily copied and would
have increased profits considerably if introduced, only one of the firms not
given the new technology had copied it.
The reason, it seems, is that the employees who would have used the new
dies (cutters and printers) were paid piece rates, that is, the employees were
paid per panel they cut. The payment method mattered because the new
technology did not speed up the process of cutting, which would have increased the pay of the cutters. Instead the cost reduction came from saving
leather, which would enhance the profits of the owners, but would not have
benefited the workers.
Because the cutters and printers did not stand to gain by saving leather,
they had no interest in adopting the new technology. This was especially the
case given the initial learning period in which the number of panels cut would
actually be lower than before, meaning the workers would, for a short period,
make less money. So they complained to their employers that the new dies
did not work very well. Owners, lacking any independent way of verifying the
competing claims of the Italo-American couple and their own cutters showed
little interest in the new technology.
Except one. One of the larger firms had a different pay system – the cutters
were paid a fixed monthly salary rather than per panel that they cut. This firm
purchased (and used) 32 of the new dies, apparently without resistance from
the cutters. As long as none of the other firms adopted the new technology,
this firm would then have been making substantial innovation rents due to the
reduced cost of materials.
If the competitive process worked in Sialkot that way economists think that it
should, then this firm should have expanded its share of soccer ball production, eventually forcing other firms to either adopt the new technology, or to
drop out. We do not know if that is what happened.
Figure 6.23: A worker at the firm that adopted
the new technology.
336
MICROECONOMICS
- DRAFT
300
30000
200
Hours of work per 10 million
lumen hours (ratio scale)
Labor hours per 100 bushels
of wheat (ratio scale)
10000
100
50
30
20
10
3000
1000
5
300
100
30
10
3
3
1
1840
1860
1880
1900
1920
1940
1960
1980
1800 1845 1855 1875 1895 1920 1940 1990 1992
Year
(a) Improvements in agricultural technology
Year
(b) Improvements in light technology
This case makes it clear that firms are made up of people, and the sometimes
incomplete information and conflicting interests among them constitute impediments to improvements that in principle at least would allow for mutual gains
to be shared among employees and owners.
6.14
Characterizing technologies and technical change
Production technologies shape how we live, and ongoing changes in technologies are revolutionizing the world. The Industrial Revolution and changes in
technology since have transformed the economies of Europe and North America from largely agricultural production to manufacturing and later servicebased livelihoods. Included were the shift of most work out of the home and
into the factory or office, the enormous increase in the scale of production of
typical firms, the widespread replacement of human labor by machines and
vast increase in the quantity of goods and services available along with a
decline in the amount of time in one’s lifetime spent working.
Figure 6.24 shows the scale of these productive improvements for two technologies: agricultural output and light. Panel a shows the change in the the
number of hours required to produce 100 bushels of wheat. In 1830, farmers
needed 275 hours to produce 100 bushels of wheat or over thirty-four 8-hour
workdays in contrast with merely 3 hours required to get the same amount of
wheat in 1987 (and even less time today).
In panel a, we show how the number of hours of labor time have decreased to
obtain 100 bushels of wheat. A bushel of wheat is approximately 60 pounds or
Figure 6.24: Improvements in farming and
lighting technology over time. In both panels,
improvements in technology show the reduced
number of hours of labor required to obtain the
indicated output. The vertical axis measures what
is called a ratio scale so that, for example, the
distance between 20 and 100 is the same as the
distance between 10 and 50 (the ratio of the first
to the second number is the same in both cases).
This is equivalent to a logarithmic scale, so the rate
of change of the measure is the slope of the lines
shown. Sources: Nordhaus (1996) and Spielmaker
(2018).
xL3
k2
xL2
k1
xL1
xCD
3
xCD
2
xCD
1
l1
l2
337
Quantity of capital goods, k
k3
Quantity of capital goods, k
Quantity of capital goods, k
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
xS1
xS2
xS3
l3
Hours of labor, l
(a) A Leontief Technology
Hours of labor, l
(b) A Cobb-Douglas Technology
about 27.5 kilograms of wheat. The graph shows the amount of time required
to obtain 100 bushels, decreasing from over 275 hours in 1830 to merely 3
hours in 1987 – a 90-fold improvement in productivity.
In panel b, we show the amount of labor to obtain 1000 lumen hours. A lumen
is a standard measure of light intensity equivalent to the light of one candle.
Hours of labor, l
(c) A perfect substitute technology
Figure 6.25: Substitutability between labor and
capital goods with different unit isoquants.
Panel a. Leontief isoquants illustrating an elasticity
R Esubstitution
M I N D E R A production
is one
of
of zero Paneltechnique
b. Cobb-Douglas
isoqaunts illustrating
an elasticity
substitution
particular
way of producing
anofoutput,
x, l, kof.
one. Panel c. three isoquants of a technology in
A production function (for example, the
which the elasticity of substitution is infinite.
Cobb-Douglas) describes a technology, that
is, a set of techniques.
The increasingly steep slope of the line in the panel b indicates an acceleration of the rate of decline in the amount of labor required to produce a given
mount of light. Panel b shows a vast increase in productivity of human labor
demonstrated by the decline in how much labor time was required to produce
10,000,000 lumen-hours of light.
Two hundred years ago, to make even one candle required immense amounts
of work to benefit from the output: a modest amount of artificial light.
In the contemporary world, though, light arrives at the flick of a switch and
the labor required to produce the electricity and the advanced technology of
super efficient light-emitting diodes (LEDs) is measured in minutes not days
or weeks. Both of these outputs – wheat and light – show the ways in which
productivity has increased over time as a consequence of human ingenuity.
You can see from the figure that the pace of productivity improvements is
accelerating in lighting (the decline in labor required is steepening over time)
and slowing down in farming.
Interpreting technological change with production isoquants
To understand how technologies impact how we work and live, and why new
technologies continue to revolutionize our economy and society, think about
five dimensions of a technology.
• Economies of scale: Does increasing all inputs by a factor of S increase
output by more than a factor of S?
M - C H E C K A production function A is more
labor-intensive than production function B if
for any given ratio of wages to the price of
capital goods, the cost minimizing choice of
inputs will be to hire more labor hours when
using A than when using B.
338
MICROECONOMICS
- DRAFT
• Overall productivity: For a given set of inputs how much output is produced?
• Input intensity: Does the production process rely more on the input of labor
(as in caring for children or the elderly, or doing scientific research), or
capital goods (as in manufacturing) or natural resources, information, or
some other input? Labor intensive and capital-goods intensive production
is illustrated in Figure 6.20.
• Complements and substitutes: If the amount of one input used increases
and this raises the marginal product of the other input then the inputs
are complement. Example: computer driven welding robots increase the
marginal productivity of the engineers who program them. If the increase
in one input used reduces the marginal product of another input, the two
inputs are substitutes. Example: computer driven welding robots reduce
the marginal product of manual welders (possibly to zero).
• Input substitutability: Must inputs be used in some fixed proportions or
can one input be substituted for another. An example of fixed proportions:
a truck needs a driver, adding a second driver or a second truck does
M - C H E C K Figures 6.20 a. and b. illustrated
the isoquants of Cobb-Douglas and Leontief
technologies, one of which is more laborintensive (less capital-intensive) than the
other. You can check the input-intensity
by looking at the slope of the isoquants
(that is the, the marginal rate of technical
substitution or the ratio of the two marginal
products) for a given ratio of the inputs (that
is, along a ray from the origin). Technology
A (shown by x1A ) is more capital goods
intensive than Technology B (shown by
x1B ) because, for a given ratio of inputs,
the marginal productivity of capital goods
(relative to the marginal product of labor) is
higher.
not add much to the transportation services delivered. An example of
substitution: calculations done "by hand" with pencil and paper can also be
done by a computer, using more capital goods and less labor.
The extent to which one input can be substituted for another in production is
termed the elasticity of substitution defined as the percentage change in the
minimum cost input proportions associated with a percent change in the ratio
of the wage rate to the price of capital goods. The elasticity of substitution
ranges from:
• zero in the Leontief production function (zero change in input proportions,
no matter how much the input prices change (illustrated in panel a of
Figure 6.25) to
• one in the Cobb-Douglas production function (if the ratio of the wage to the
price of the capital good doubles, the ratio of amount of the capital good
to the amount of labor used by a cost minimizing producer will double,
illustrated in panel b of Figure 6.25), to
• infinity where the inputs are called perfect substitutes, and using one of
the two inputs, but not both will generally be cost minimizing (illustrated in
panel c of Figure 6.25).
The elasticity of substitution tells us how linear the isoquants are, ranging
from perfectly linear (perfect substitutes), to extremely kinked (the Leontief
case).
Examples of how these dimensions of technologies altered ways of life in the
ELASTICITY OF SUBSTITUTION is the percentage change in the minimum cost input
proportions associated with a percent
change in the ratio of the wage rate to the
price of capital goods.
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
past, and continue to do so today are given in Table 6.3. Key questions about
technology today include:
• Are capital goods in the form of robots substitutes for workers doing routine tasks and at the same time complements of engineers who operate
them? If so, the marginal products of the two kinds of workers will diverge,
possibly generating greater inequality between engineers and routine task
workers.
• As the ratio of capital goods to labor input continues to increase, this may
depress the marginal product of capital goods and raise the marginal product of labor as would be expected if capital goods and labor are complements. This could be the basis of greater income equality between owners
of capital goods and the workers they employ. But by how much?
• Sectors of the economy with labor intensive production functions – such
as education, security services, entertainment, child and elder care, and
health services – are increasing their share of the economy at the expense
of capital goods intensive sectors such as manufacturing. Will this result
in greater scarcity of labor relative to capital goods and an increase in
workers’ bargaining power?
• Will devices and algorithms associated with artificial intelligence become a
substitute or the work of engineers and other professionals, driving down
their marginal products?
6.15
Conclusion
Economics (as you read at the beginning) is the study of how people interact
with each other and with our natural surroundings in producing and acquiring
our livelihoods. The technologies studied in this chapter – summarized mathematically by production functions– describe how we can produce our livelihood
by transforming nature – crops mineral resources and energy – in order to
provide the goods and services that make up our standard of living.
The available technologies and the ways owners of firms seek to maximize
their profits by choosing techniques of production that minimize their costs
of production have important effects on how we interact with each other in
this process (the other part of the definition of economics). If technologies
are highly efficient, then people will have the opportunity for high living standards for all. But if there are important economies of scale in production then
it is likely that the economy will be dominated by a limited number of large
firms whose owners may be able very disproportionate share of the high levels of production. If labor is highly productive relative to capital goods, then
those who own the capital goods will not be able to earn such high incomes
as those – virtually everyone – who provide labor to the system of produc-
339
340
MICROECONOMICS
Characteristic
- DRAFT
Example:
Example from Industrial Revolu-
Cobb-Douglas
tion
Examples from today and future
Economies of
x = ql a kb
a + b >
Industrial revolution increased
New technologies (e.g. 3-D print-
scale
economies of
economies of scale leading to
ers) may reduce economies of
scale a + b < 1
larger firms
scale but large first copy costs
1
diseconomies of
(“prototyping”) imply economies
scale
of scale (e.g. R&D for producing a
drug)
Overall productiv-
q
ity
Increases in productivity allowed for
Is a long term slowdown in produc-
improved living standards including
tivity growth in our future?
less work
Labor intensity
a
a fell and the ratio of capital goods
Labor with engineering and net-
to labor input rose
working skills may be replacing both
capital goods & other labor
Substitutes or
Inputs are com-
Capital goods were substitutes for
Artificial intelligence (AI) may be-
complements
plements
some kinds of labor (manual, rou-
come a substitute for even highly
tine) and complements for others
trained engineering and other labor
(engineering, design)
Elasticity of sub-
Elasticity of sub-
For many early machining and pro-
If the elasticity of substitution is low,
stitution
stitution = 1
duction line processes, substitution
then the continuing increase in the
was very limited.
quantity of capital goods per worker
could allow wages to rise relative to
profits.
tion.
But technologies and the division of labor that results when people specialize
according to comparative advantage is just one part of economic knowledge.
Equally important are the wants and needs of people and how these are expressed in our willingness to pay for goods when they are supplied in markets.
We turn to market demands next.
Making connections
Economies of scale and learning by doing are among the main reasons for
the the division of labor and specialization, which makes important contributions to human well being; but we will see in Chapter 8 that economies
of scale and learning by doing a may also limit the degree of competition in
markets.
Markets as a means of coordination: The opportunity to exchange of goods
expands the set of feasible outcomes available to people and nations by
providing facilitating specialization and the division of labor.
Table 6.3: The ways in which technologies differ
and why it matters.
P R O D U C T I O N : T E C H N O L O G Y A N D S P E C I A L I Z AT I O N
341
External effects, coordination failures and poverty traps: The positive external
effects associated with economies of agglomeration result in many possible
Nash equilibrium patterns of specialization; countries may specialize in
goods that keep them poorer than had they specialized in some other way.
Constrained optimization: the choice of technology Minimizing cost subject to
a production function constraint and maximizing utility subject to a budget
constraint have many features in common. They are both examples of
maximization (or minimization) under constraints.
Innovation rents A firm that succeeds in finding a new technology that lowers
costs of production at existing input prices can make substantial economic
profits, called innovation rents, until others adopt the same or similar innovation.
Important ideas
specialization
production possibilities frontier
diversification
technique of production
production function
production isoquant
division of labor
average product
marginal product
economies of agglomeration
marginal rate of transformation
relative price
cost minimization
equalization of marginal products and input prices
marginal rate of technical substitution
economies of scale
constant returns to scale
diseconomies of scale
wages
isocost line
rental price of capital goods
diminishing marginal productivity
short run/long run
technical efficiency
opportunity cost of capital
342
- DRAFT
MICROECONOMICS
Mathematical Notation
Notation
Definition
x, y
x̄, ȳ
p
l
k
f ()
al
ak
a
b
q
x
S
w
pk
c
goods produced using labor and capital (or just labor)
maximum feasible production of goods x and y, given the current technology
price
firm input: labor
firm input: capital
production function
required amount of labor inputs in Leontief production function
required amount of capital goods in Leontief production function
intensity of labor in Cobb-Douglas production
intensity of capital in Cobb-Douglas production
parameter of productivity, Leontief and Cobb-Douglas production
isoquant, a fixed amount of good x that can be produced by different combinations of labor and capital
scale factor for increases in all inputs
wage
cost of renting capital goods
cost function
Note on superscripts: L: related to labor; K: related to capital.
Discussion Questions
See supplementary materials.
Problems
See supplementary materials.
Works Cited
See reference list.
7
Demand: Willingness to pay and prices
DOING ECONOMICS
This division of labour... is the necessary... consequence of...the propensity to
truck, barter, and exchange one thing for another. It is common to all men, and
to be found in no other race of animals... Nobody ever saw a dog make a fair
and deliberate exchange of one bone for another with another dog.
Adam Smith, The Wealth of Nations, Book 1 ch 2
Ancona is a town on the Adriatic coast of Italy southeast of Venice. It hosts
one of the many daily fish markets that sell to European restaurants and fish
dealers.
Because fish (notoriously) spoil rapidly even with refrigeration, the price of fish
on any one day depends largely on the amount of fish brought to the market
that day (since none can be carried over from previous days). Economists
view fish markets as a kind of ideal experiment for studying how supply and
demand determine the prices at which goods are bought and sold .
Figure 7.2 shows that the average daily prices and the average daily quantities of fish sold in the Ancona market:
• if the price per kilogram of fish is high, the quantity of fish bought and sold
is less, and
• if the price per kilogram of fish is low, more kilos of fish are transacted.
One explanation for the downward-sloping line in the figure summarizing this
relationship is that typical buyers in the Ancona fish market will buy more
This chapter will enable you to:
• Apply constrained utility optimization
to the the problem of demand, relating
a person’s willingness to pay to their
purchases of goods.
• Derive a person’s demand curve from a
utility function describing the person’s
preferences.
• Understand that consumption is often
a social activity, so our preferences for
particular goods (for example smoking
tobacco products) often depend on what
others are consuming.
• Explain how people change their purchases when prices or income change.
• Understand how these responses reflect
both income and substitution effects and
use these concepts to explain the effects
of a proposed carbon tax and citizen
dividend.
• Use the concept of consumer surplus
and understand the conditions under which it makes sense to sum the
consumer surplus of many people.
• Explain how market demand curves can
be be derived from individual demand
curves.
• Use the price elasticity of demand to
explain the effects of price increases for
example resulting from policies such as
placing a tax on sugary drinks.
fish if the price is lower. Another is that the greater quantity of fish brought to
market on any given day, the lower will be the average price per kilogram of
fish.
To understand how the price and quantity of fish purchased is determined,
an essential concept is demand, as measured by the amount a person is
willing purchase at any given price. For fish and other goods, knowing how
Figure 7.1: Mosquito nets save lives; how widely
used they are depends on the price. See Figure
7.3.
Daily average price per kilogram (Euros)
344
MICROECONOMICS
12
- DRAFT
●
10
●
●
●
●
●
●
●
8
●
●
●
●
●
●
●●
●
6
●
0
Figure 7.2: Prices and quantities of fish bought
and sold in the Ancona market. The plot of
daily average prices of fish in the Ancona market
against the same day’s quantity of fish sold can be
summarized by a downward-sloping curve.
●
100
●
●
●
●
●
●
●
●
●●
●
●●
●
● ●
●
●
●
●
●
●●
●
● ●
●●● ●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●●
●
●
●
●
●
200
300
400
●
●
●
●
●
●●●
●
●
●
●
●
500
Daily quantity (kilograms)
the amount purchased depends on the price is also an important piece of
information in the design of economic policies.
Here is an example. There are now many low-cost life-saving preventative
health products such as insecticide-treated mosquito nets, tablets to eradicate
parasitic stomach worms, and water purification products. In many countries
in Africa, Asia and Latin America, these products prevent illness and death of
their users, and also limit the spread of infectious diseases to others, but they
are used sparingly if at all. Some policy-makers think these products should
be provided free of charge to low-income families to encourage the use of the
products.
Other policy-makers disagree, suggesting that there should be a cost to acquiring these products to discourage wasteful use through better targeting of
who gets the products. The question then arises: how will the take-up of the
products depend on the price? Will charging even a small price significantly
discourage use?
Economists have conducted experiments in 8 countries to find answers to
these questions. In the experiments, potential users are randomly selected to
be offered the goods free or at one or more different prices. The average use
of the products at each price (including zero) is then recorded. Some of their
results are shown in Figure 7.3.
Figure 7.3 shows that the effect of charging higher (even if very low) prices
F AC T C H E C K Adam Smith’s claim in the
head quote that humans are unique among
all animals in our division of labor and
exchange of goods is probably right about
dogs. But Smith is certainly wrong about the
many other species such as ants and other
social insects that practice a very advanced
division of labor and specialization. Different
species of fish exchange services in what
are termed "biological markets."
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
$6
Bednets (Kenya 2007, pregnant women)
Bednets (Kenya 2007)
Cement latrine slabs (Tanzania 2015)
Clorin (Zambia 2006)
Deworming (Kenya 2001)
Plastic latrine slab (Tanzania 2015)
Soap (Guatemala 2009)
Water filters (Ghana 2009)
Price in USD
$5
$4
345
Figure 7.3: The demand for preventative health
products: Take-up rates at various prices
and when available for free. Our measure of
demand is the take up rate, that is the fraction of
the population that acquires the product (whether
free or for a price). For most products the demand
is substantially less when even a small price is
charged compared with when the good is available
for free; and higher prices are also associated
with substantially less demand than lower prices.
Source: Dupas and Miguel (2017).
$3
$2
$1
0.
0%
10
.0
%
90
.0
%
80
.0
%
70
.0
%
60
.0
%
50
.0
%
40
.0
%
30
.0
%
20
10
.0
%
$0
Household Take−up Rate
was to reduce the amount of the product used, in some cases by a substantial
amount.
• In Zambia, for example, increasing the price of a disinfectant tablet from
nine cents to twenty-five cents reduced the fraction of the population using
the product from 76 to 43 percent.
• Only 43 percent of a group of pregnant Kenyan women purchased insecticide treated mosquito nets when the price was 60 cents; virtually all used
the nets when they were provided without charge.
• A program in Kenya that had initially given away de-worming tablets for
children, but later introduced a charge of thirty cents per child found that
usage of the tablets fell from 75 per cent of the affected population to just
18 percent.
On the basis of this information, the Poverty Action Lab at MIT, led by economics Nobel Laureates Abhijit Banerjee and Ester Duflo, suggested that
there are good reasons to make these products available without charge or
highly subsidized to ensure very low prices.
We begin our analysis of how people spend their money – whether on fish
or mosquito nets – with a basic fact: there are limits to how much a family or
person can spend.
R E M I N D E R Economists have researched
preferences based on observing people’s behavior in real situations and in experiments.
We reviewed some of the key findings from
this work in Chapter 2 and we will review
more experimental data in later chapters.
346
MICROECONOMICS
- DRAFT
Figure 7.4: Budget constraint for coffee and
data. The budget set is shaded in green and
the budget constraint is the dark green line
on the border of the budget set. Consumption
bundles (x, y) in the budget set and on the budget
constraint can feasibly be obtained with the current
budget (m) at going market prices for x and y,
px and py . Outside the budget constraint, in the
shaded blue area, the bundles of x and y cannot
feasibly be obtained at going market prices with the
existing budget.
m
y=
py
Gigabytes of data, y
Infeasible
(outside the budget)
Feasible
(within the budget)
Budget Constraint
m ⎛px⎞
y = − ⎜ ⎟x
py ⎝py⎠
Kilograms of coffee, x
x=
m
px
7.1 The budget set, indifference curves and the rules of the game.
To understand how prices influence the take up of one of the life saving health
products in Figure 7.3 or the amount of some good that we will consume, think
about someone who has a total amount of money to spend m that she has in
cash, savings, available credit, and so on.
The budget constraint
We shall consider a person, Harriet, and the decisions she needs to make. A
person’s budget set states what bundles (x, y) are feasible for her to consume
given her budget and market price of the goods:
m
px x + py y
m
Prices of Goods ⇥ Quantities of Goods Purchased
(7.1)
Expressing this inequality as an equality – assuming that Harriet would not
consume less than her budget allowed – we have the budget constraint
m
Budget
=
px x + py y
= Prices of goods ⇥ Quantities of goods purchased
(7.2)
(7.3)
Equation 7.3 is a statement about prices and budgets. But it is also a statement about the rules of the game and preferences. They are as follows:
R E M I N D E R We saw budget constraints in
Chapter 3: the slope was a price ratio that
represents the opportunity costs of spending
money on one good in terms of how much
less of the other you can afford depending
on your budget.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
347
• No gifts, thefts or consumption as a matter of citizen rights: You can consume only what you pay for; so no gifts or goods provided by government,
or acquired by theft.
• No altruism or concerns about environmental sustainability: You consume
the most that your budget allows and you consume it yourself, rather than
giving it to or sharing it with others.
We can rearrange the budget constraint, Equation 7.8, to obtain a line we can
draw on the x and y-axes we use for indifference curves:
y=
m
py
px
x
py
(7.4)
B UDGET S ET & B UDGET C ONSTRAINT The
budget set is the set of all feasible purchases
of the bundles of x and y with current budget
m, such that m px x + py y. The budget
constraint is the border of the budget set
showing all combinations that exhaust the
budget, i.e. for which the constraint holds
with equality, m = px x + py y.
We plot the budget constraint in Figure 7.4. Examining the two terms on the
right-hand side of Equation 7.4, we can see that if Harriet were to consume
only good y and no good x, then she would consume pm units of good y which
y
is the intercept of the budget constraint with the y-axis. As Harriet buys more
of good x, she moves along the budget constraint with the slope
px
py
indicat-
ing the rate at which she can sacrifice good y for good x. If she were to buy
only good x, she could afford x = pm units of good x.
x
The negative of the slope of the budget constraint is a relative price and
measures the opportunity cost of obtaining good x in terms of the amount
of good y that Harriet must sacrifice because her funds are limited. The
(negative of the) slope of the budget constraint is another marginal rate of
transformation, it tells us the terms on which a reduced amount of good y can
R ELATIVE P RICE A relative price shows
the price of one or more goods relative to
another good, as such it is indicated by a
ratio of the one price relative to the other.
Relative prices show the opportunity cost
of having more of one good in terms of the
lesser quantity of the other imposed by the
budget constraint.
be "transformed into" additional amounts of good x while just satisfying the
budget constraint.
M-Note 7.1: Budget for coffee and data
For particular values of m, px and py we can graph the budget constraint. Consider the
following example:
• Harriet has a budget (m = $50) to spend on kilograms of coffee, x, and gigabytes of
data, y.
• The price of a kilogram of coffee, px , is $10.
• The price of a gigabyte of data, py , is $5
Putting these pieces of data together, therefore, the budget constraint is 50 = 10x + 5y.
We can re-arrange the budget constraint as we did in Equation 7.4:
y
=
y
=
50 10
x
5
5
10 2x
(7.5)
Equation 7.5 is a line with an intercept at pm = 10 on the y-axis, an intercept of 5 on the xy
axis and a slope of p =
2. Such a curve would look like bc1 in Figure 7.5.
M - C H E C K From the budget constraint, we
xpx
know: y = pm
py . We can take the first
y
derivative to see that the slope of the budget
constraint is:
dy
px
=
dx
py
That is the opportunity cost of x in terms of y.
348
MICROECONOMICS
- DRAFT
Figure 7.5: Utility-maximizing consumption
bundle. Harriet maximizes her utility subject to her
budget constraint bc1 . At point a she consumes
too little of x and too much of y (her marginal utility
of y is much lower than her marginal utility of x, or
her mrs(x, y) is too high, and she would be better
off if she consumed less y and more x. Conversely,
at c, she consumes too little of y and too much of
x (her marginal utility of x is much lower than her
marginal utility of y, or her mrs(x, y) is too low, and
she would be better off if she consumed less x
and more y. She maximizes her utility at b where
her marginal rate of substitution, mrs(x, y) = uux ,
m
y=
py
Gigabytes of data, y
a
mrs = mrt
y
yb
equals her marginal rate of transformation or the
p
price ratio of x to y, mrt (x, y) = px .
u3
b
y
u2
c
u1
Budget constraint, bc1
xb
x=
Kilograms of coffee, x
m
px
Checkpoint 7.1: Sketching a budget constraint
a. Consider two goods: vegetables (x) which have a price of 4 euros per kilogram and meat (y), which has a price of 10 euros per kilogram. You have a
budget of 50 euros a week for meat and vegetables for your family. Sketch
your budget constraint.
b. The price of vegetables increases to 5 euros per kilogram. What happens to
your budget constraint? Sketch and explain.
Budget constraints, indifference curves and the amount demanded
In Figure 7.5 we show three of Harriet’s indifference curves. Remember at
any given point on the indifference curve, the negative of its slope tells how
much Harriet values the good on the x-axis compared to her valuation of the
good on the y-axis, that is, her marginal rate of substitution (mrs(x, y)). Her
marginal rate of substitution is her willingness to pay to get more of good x,
namely how much of good y she would be willing to part with, in order to get
one more unit of good x. So,
(Negative of) the slope of an indifference curve = mrs =
ux
uy
Harriet wants to get to the highest indifference curve that she can, given her
budget. This is the point at which the budget constraint is tangent to her highest attainable indifference curve. For the two curves to be tangent, the slope
of her indifference curve must equal the slope of the budget constraint. Or,
R E M I N D E R We reason here in the same
way we did in Chapter 3 that the utilitymaximizing choice is the point of tangency
between the highest attainable indifference
curve and the feasible frontier or budget
constraint.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
349
the marginal rate of substitution must equal the marginal rate of transformation.
mrs(x, y) =
ux
uy
px
= mrt (x, y)
py
=
(7.6)
This is the principle of constrained utility maximization that you learned in
Chapter 3 because it determines how much of a good people will demand
or want to buy at given prices. When Harriet uses the mrs = mrtrule, her
trade-offs of one good for another in terms of utility (mrs(x, y)) equal the
opportunity costs of the two goods in terms of each other (mrt (x, y)), where
the opportunity costs are given by their prices. Remember that money she
spends on one good means money she cannot spend on another good:
capturing the essential idea of an opportunity cost.
Equation 7.6 – the mrs = mrt rule – states that the relative valuation of x and
y along her indifference curve must equal the opportunity cost of x in terms of
y along her budget constraint.
The point where Equation 7.6 holds, is Harriet’s utility-maximizing consumption bundle or the total quantity of goods x and y that Harriet will buy. A useful
interpretation of the marginal rate of substitution occurs when good y is not
data, but instead the amount of money left over from the budget after purchasing good x. In this case, the marginal rate of substitution can also be thought
of as Harriet’s willingness to pay for the good x, the amount of money she
will pay for an additional unit of x when she has already bought the quantity
x.
We need to distinguish between two concepts derived from the mrs = mrt
rule:
• Quantity demanded: The point (x, y) (the consumption bundle) at which
p
Harriet’s mrs(x, y) = px are her quantities demanded, these are the
y
amounts of each good that she consumes at the given prices and given
income.
• Demand function: Using the rule, we can derive a demand function for
each good, where the quantity demanded depends on income (m) and
prices ( px and py ).
M-Note 7.2: The mrs = mrt rule applied to demand for goods
The problem the person faces is to maximize their utility subject to their budget constraint:
Vary x and y to maximize
u(x, y)
subject to the constraint
y
(7.7)
=
m
py
Substituting Equation 7.8 into Equation 7.7, the problem becomes:
Vary x to maximize
u(x,
m
py
x
px
)
py
px
x
py
(7.8)
W ILLINGNESS TO PAY A person’s willingness
to pay for a good x in terms of y (for example
budget left over to buy other goods) is their
marginal rate of substitution between y and x
when they are already purchasing the bundle
(x, y).
350
MICROECONOMICS
- DRAFT
Taking the partial derivative of the utility function with respect to x, the first order condition
for maximum utility is:
∂u
∂x
ux
)
uy
mrs(x, y)
px
uy = 0
py
=
ux
=
px
py
=
mrt (x, y)
In the constrained utility maximum is a set of purchases such that the marginal rate of
substitution is equal to the relative prices or the marginal rate of transformation.
7.2 Income, prices and offer curves
To study how prices and incomes (budgets) affect the demand for goods we
ask a hypothetical "what-if" question: how much of good x would someone
purchase if her budget were m and the price of good x were px and the price
of good y were py .
A demand function shows the quantity purchased of x that results for the
various values of the prices of both goods and the budget, px , py and m. So
x(m, px , py ) is the demand for x as income (m) or the price of x ( px ) and the
price of y( py ) change. We use the term demand curve when we refer to the
simpler 2-dimensional graphical relationship x( px ) where we see how the
amount purchased varies its price ( px ) varies holding constant all of the other
influences on the demand for x.
D EMAND FUNCTION , DEMAND CURVE A
demand function provides an answer a
hypothetical "what-if" question: how much
of good x would a person purchase if her
budget were m, price of goods x and y
were px and py ? A demand curve is a 2dimensional graphical representation of a
demand function showing the purchases
of x that result for the various values of px
(with the other influences on demand held
constant).
I NVERSE
DEMAND FUNCTION , INVERSE
DEMAND CURVE The
We sometimes use a demand curve in which, instead of quantity sold depending on the price x = x( p), price depends on the quantity sold, p = f (x).
This is called an inverse demand curve based on the inverse demand
function, because it is the mathematical inverse of the conventional demand
function.
The inverse demand function contains exactly the same information and the
demand function and the inverse demand curve looks identical to the conventional demand curve (it is downward-sloping). What differs is the hypothetical
question for which the inverse demand function provides an answer. Instead
of asking how much of a good will be purchased at a given set of prices and
a budget, the inverse demand function answers the question: if the budget
and the price of the other good are m and py what is the maximum price
px that the buyer would be willing to pay to purchase an amount x of the
good?
A change in income: The income-offer curve
To understand these changes, therefore, we examine Figures 7.6 a. and 7.6
b. As Figure 7.6 a. shows, as Harriet’s income changes, her budget constraint
shifts. That is, the intercept with both axes, pm and pm , shifts up as income (m)
y
x
inverse function answers
the hypothetical question: what is the highest
price that a person be willing to pay in
order to purchase a given amount of some
good, given her budget and the prices of
other goods? The inverse demand curve
is the simplified 2-dimensional graphical
representation of this function as px = f (x).
m3
py
m3
py
m2
py
m2
py
Gigabytes of data, y
Gigabytes of data, y
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
m1
py
bc1
bc2
u1
u3
u2
Income−offer
curve
c
m1
py
b
a
bc3
bc1
m1 px
m2 px
351
m3 px
x = m 1 px
Kilograms of coffee, x
bc2
x = m 2 px
bc3
x = m 3 px
Kilograms of coffee, x
(a) Budget constraint & shifts in Income
(b) The income-offer curve
goes up and shifts down as her income goes down. Consider the three budget
constraints in Figure 7.6 a where only income changes, but the prices of the
two goods do not change.
m
m
• Status quo: She starts with an income of m2 with intercepts p 2 and p 2 .
y
x
• Income decrease: If her income decreases to m1 , then the intercepts of her
m
m
budget constraint shift downwards and to the left to p 1 and p 1 , so she can
y
x
buy less of both goods.
• Income increase: If her income increases to m3 , then the intercepts of her
m
m
budget constraint shift upwards and to the right to p 3 and p 3 , so she can
y
x
by more of both goods.
Considering different levels in Harriet’s income we can superimpose Harriet’s
indifference curves to find the consumption bundle for each income level that
would maximize Harriet’s utility using the mrs = mrt rule. The path traced
out by the points (x, y) as m increases is called her income-offer curve. Her
income-offer curve is also called her expansion path because it shows the
effect of expanding her feasible set (by increasing her budget). In Figure 7.6
b, her income-offer curve is upward-sloping, showing the effect of an increase
on her income on her consumption of the goods, x and y. As she gets more
income, she would consume more of both goods.
The income-offer curve allows us to understand the types of goods people
consume.
• Normal goods: normal goods are goods like coffee and data where people buy more as their income increases, or less of them as their income
decreases.
• Inferior goods: inferior goods are goods like cheap staples, like white
Figure 7.6: Harriet’s budget constraint with
shifts in income & her income-offer curve. In
panel a Harriet’s budget constraints with three
levels of income are shown (m1 , m2 and m3 ) with
the corresponding budget constraints bc1 , bc2
and bc3 shifting outwards as income increases.
In panel b Harriet’s budget constraints are shown
tangent to three indifference curves, u1 , u2 , and
u3 . The points where they are tangent are where
mrs(x, y) = mrt (x, y). The curve joining all the
points at which Harriet maximizes her utility as her
income changes illustrate her income-offer curve.
I NCOME - OFFER CURVE The income-offer
curve is the path traced out by the points
(x, y) that maximize the decision-maker’s
utility as money-income, m, increases,
holding the price of good x, px , and the price
of good y, py constant.
352
MICROECONOMICS
- DRAFT
less of inferior goods as their income increases and more of them as their
income decreases.
Figure 7.7 shows a situation in which Harriet’s income increases, but her
consumption responses for the two goods differs. For good y, on the vertical
axis, Harriet consumes more of it as her income increases from m1 to m2 : she
Quantity of normal good, y
sandwich bread, basic rice, or instant noodles: people tend to consume
h
y3
g
y2
bcm3
f
y1
bcm1
increases her consumption from y1 to y2 . For good x, on the horizontal axis,
on the contrary, Harriet consumes less as her income increases from m1 to
m2 : she decreases her consumption from x2 to x1 as her income increases.
As a result, she has a downward-sloping income-offer curve.
Checkpoint 7.2: Inferior indifference curves
1. On your own set of axes, re-draw Figure 7.7. What condition must be true at
each of points f, g and h?
2. Add the relevant indifference curves to your figure. What do you think they
look like? Explain.
A change in prices: the price-offer curve
As prices increase, Harriet has to reduce her purchases of at least one good,
and possibly both goods because she’ll have less money to spend on the
goods if they become more costly. As the price of a good changes, holding
the price of the other good constant, the slope of the budget constraint will
change. For example, if the price of good x changes while the price of good y
remains the same, the budget constraint will pivot around the y-axis intercept
depending on whether the price of x increases or decreases. In Figure 7.8 a.
on the left-hand side we show budget constraints based on three prices for
good x:
• An initial price along budget constraint bc2 .
• A price increase, which steepens the slope of the budget constraint to bc1 .
• A price decrease, which flattens the slope of the budget constraint to bc3 .
At the point of tangency of each budget constraint with a corresponding indifference curve in panel b, we can see how Harriet would respond to a change
in the price of goods x and y, other things equal. The path traced out by the
number (x(m, px ), py ) as px changes (holding the budget, m and py constant)
is called her price-offer curve. We illustrate the price-offer curve in Figure 7.8
b.
Points a, b and c, all map her utility-maximizing choices and the curve that
connects these points is the price-offer curve as we introduced in Chapter
3.
bcm2
Income−offer
curve
x1
x2
x3
Quantity of inferior good, x
Figure 7.7: Inferior Goods and An Increase in
Income. The downward-sloping income-offer curve
for an inferior good (x) and a normal good (y).
As the person’s budget increases, they consume
less of good x, showing that x is an inferior good
(decreasing from x3 to x2 to x1 ). Consumption of
good y increases as income increases, showing
that y is a normal good (increasing from y1 to y2 to
y3 ).
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
Price−offer
curve
m
Money left over, y
Money left over, y
m = 10
bcpx3
px3 = 1
bcpx2
px2 = 0.5
m
= 10
px3
c
y3
u2
a
y1
m
= 40
px1
x1 = 6
u1
x2 = 9 x3 = 10.5
Kilograms of fish, x
(a) Budget Constraint & changes in the price of x
(b) The price-offer curve
7.3 Cobb-Douglas utility and demand
You already encountered Cobb-Douglas utility in Chapter 3. We build on that
base and explore a person’s choice of her utility-maximizing consumption
bundle using indifference curves based on a Cobb-Douglas utility function
and budget constraints. The Cobb-Douglas utility function has the following
With a and 1
bc2
px = 0.5
bc1
px = 1
Kilograms of fish, x
general form:
u3
bc3
px = 0.25
b
y2
bcpx1
px1 = 0.25
m
= 20
px2
353
Figure 7.8: Harriet’s budget constraint with
changes in price & her price-offer curve.
In panel a Harriet’s budget constraints for
three prices for fish, good x, px1 , px2 and px3 are
shown with the corresponding budget constraints
pivoting inwards for a price increase (such as
for bc px2 to bc px3 ) or pivoting outwards for a
price decrease (such as for bc px2 to bc px1 ). The
other good, y, is money for other goods she can
purchase. The intercept for good y is pm . Because
y
u(x, y) = xa y1
a
(7.9)
a indicating the relative strength of preference for the two
goods and the intensities sum to 1.
Using Cobb-Douglas utility, we can illustrate indifference curves for each good
as the price of the good changes and we can derive a demand curve for each
good. In Figure 7.9, we show the indifference curves for two goods: kilograms of coffee (x) and gigabytes of data (y). The marginal rate of substitution
is:
ux
mrs(x, y) =
uy
=
✓
a
1
◆⇣ ⌘
y
a
x
(7.10)
The mrs = mrt rule, then requires that,
mrs(x, y) =
✓
a
(1
◆⇣ ⌘
y
a)
x
=
px
= mrt (x, y)
py
(7.11)
From this relationship and the budget constraint as we show in M-Note 7.4, we
can derive a demand curve for each good. The demand function for kilograms
of coffee, good x is:
Demand
x(m, px ) =
am
px
(7.12)
the price of money for other goods is py = 1,
this simplifies to m as in Figure b. Note the x-axis
range differs in panels a and b. In panel b Harriet’s
budget constraints are shown tangent to three
indifference curves, u1 , u2 , and u3 . The points
where
R E M I Nthe
D Ebudget
R Theconstraint
marginaland
rateindifference
of substicurves
tangent
arefor
where
the marginal
rate of
tutionsare
means
that
a given
ratio of the
substitution (mrs(x, y)) equals the marginal rate of
quantities
of the two goods x and y making
transformation (mrt (x, y)). The curve joining all the
up
a bundle,
much
the person
val-as the
points
at whichhow
Harriet
maximizes
her utility
price
good xincrement
changes isinher
price-offer
curve.
ues aofsmall
the
amount of
x
compared to how much they value a small
increment of y is given by the ratio of the
exponents. For example, if a = 0.75, and
x = y, then the marginal utility of x is three
times the marginal utility of y (because
a/(1 a ) = 3).
354
- DRAFT
MICROECONOMICS
Gigabytes of data, y
m
= 12
py
a
6
b
c
Price−offer curve
u3
x
u2
bc1
px = 3
2
u1
bc3 px = 1
bc2
px = 1.5
4
6
8
12
p
a
Price of coffee, px
p=3
b
p = 1.5
c
p=1
0
2
4
6
Demand curve
for coffee
8
12
Kilograms of coffee, x
Equation 7.12 shows a relationship between quantity demanded (x) and price
( px ) such that the quantity demanded decreases as the price increases, or the
quantity demanded increases as the price decreases.
Rearranging equation 7.12 as x·mx = a with a Cobb-Douglas utility function,
a person implementing the mrs = mrt rule will spend a fraction of their total
budget on x, that is:
• equal to the exponent of x in the utility function namely, a a constant, and
therefore is,
• independent of the price of x and the price of y.
The fraction of the budget spent on the other good is also independent of
changes in the price of x. So, because the budget m has not changed, the
amount spent on y will also remain the same. Because the price of the
other good has not changed, the amount of good y purchased is also un-
Figure 7.9: Harriet’s price-offer curve & demand
curve for coffee. In the upper panel we show
three of Harriet’s budget constraints corresponding
to three different prices: p = 3 for bc1 , p = 1.5
for bc2 and p = 1 for bc3 . At points a, b, and c
tangent to three indifference curves, u1 , u2 , and
u3 The curve joining all the points at which Harriet
maximizes her utility for different prices of good x is
her price-offer curve.
In the lower panel, we see how the use of the
mrs = mrt rule in the top figure is the basis of the
demand curve. The prices shown on the vertical
axis are the (negative of the) slopes of the budget
constraints in the top figure. We have assumed that
a = 0.5, and that her budget for coffee and data is
$100.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
changed.
Thus, With Cobb-Douglas utility for a given price of y, py and income (m), the
only thing that differs with different prices of x, is the quantity demanded of
good x. This is why the price-offer curve in Figure 7.9 is horizontal.
Similarly the amount spent on x does not depend on the price of y, which you
can confirm from the fact that py does not appear in the demand function for
x. These are not general features of demand functions, they are specific to the
Cobb-Douglas utility function.
M-Note 7.3: Marginal rate of substitution, Cobb-Douglas utility function
Consider a Cobb-Douglas utility function:
u(x, y)
x a y1
=
a
The marginal utility with respect to of each good is:
∂u
∂x
∂u
uy =
∂y
ux =
=
axa
=
(1
1 1 a
y
a ) xa y
a
Therefore, the marginal rate of substitution of x with respect to y is:
mrs(x, y) =
ux
uy
=
axa 1 y1 a
( 1 a ) xa y a
(7.13)
Note that:
xa 1
xa
y1 a
y a
and
=
1
x
=
y
Thus, the marginal rate of substitution (Equation 7.13) becomes:
mrs(x, y) =
ux
uy
=
✓
a
(1
a)
◆
x
y
The inverse demand function
We can re-arrange the function and find the inverse demand curve. The
inverse demand curve is:
Inverse demand = px (x, m) =
am
amount spent on x
=
= price
x
amount of x purchased
Here, we have a downward-sloping demand curve where price decreases as
the quantity demanded increases. We show this demand curve in the lower
panel of Figure 7.9.
Checkpoint 7.3: Changes to prices or income with Cobb-Douglas utility
Harriet buys coffee and cookies to fuel herself while running her business. Her
355
356
MICROECONOMICS
- DRAFT
utility function for cookies (x) and cups of coffee (y) is given by the following
utility function:
u(x, y) = x0.6 y0.4
(7.14)
We assume that Harriet has a weekly budget of $10 to spend on coffee and
cookies, where the price of a cup of coffee is $3 and the price of a cookie is
$0.50.
a. Sketch Harriet’s indifference curves, her budget constraint, and calculate her
utility-maximizing consumption bundles of cookies and coffee.
b. Assume that the price of cookies now increases to $2 per cookie. What
would happen to the quantity of cookies that Harriet demands?
M-Note 7.4: Cobb-Douglas Demand Functions
The Cobb-Douglas utility function is:
u(x, y) = xa y1
a
where 0 < a < 1. The individual maximizes this function subject to a budget constraint:
m
px x + py y
=
(7.15)
Remember that the negative of the slope of the indifference curve is the marginal rate of
substitution and the negative of the slope of the budget constraint (which is also the ratio
of the prices of the two goods) is the the marginal rate of transformation. We found in
M-Note 7.3 the following:
mrs(x, y)
✓
=
a
1
a
◆
y
x
So the utility-maximizing bundle that implements the mrs
following equation:
mrs(x, y) =
✓
a
1
◆
y
x
=
px
py
) py y
=
px x
a
(7.16)
mrt rule must satisfy the
=
✓
a
1
a
◆
(7.17)
To find the demand function, substitute Equation 7.17 into the budget constraint, Equation
7.15, to isolate a value for x, which we then use to find y:
m
=
m
=
) x(m, px , py )
=
px x + px x
(a + 1
Substitute px x into Equation 7.17 to find py y and y:
py y = (1
a )m
(1
a )m
py
y(m, px , py ) =
a
1
a
a ) px x
a
am
px
✓
◆
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
We have therefore found the demands (x(m, px , py ), y(m, px , py )) as functions of the
budget, m, and the prices of the goods, px and py , given the intensity of preferences for
the goods, a . Notice that the demand for each good is independent of the price of the
other good. The demand function for each good in terms of its own price is a hyperbola.
Checkpoint 7.4: Income-offer and price-offer curves
1. Sketch the income-offer curve for Cobb-Douglas utility. Draw three CobbDouglas indifference curves with three income levels and sketch the corresponding income-offer curve.
2. The price-offer curve for Cobb-Douglas utility is a horizontal line in (x, y)
space. Make sure that you can sketch a similar figure to the one we used in
Figure 7.9. Notice, though, that in that figure a = 1
a = 0.5. In Chapter
3 we used a = 0.6 and 1
a = 0.4 to consider different preferences for
Living (x) and Learning (y). How would these different exponents change the
slopes of the indifference curves? Sketch approximately the corresponding
price-offer curve for two goods, coffee (x) and cookies (y) for three different
prices of coffee.
7.4 Application. Doing the best you can dividing your time
We can apply the Cobb-Douglas utility function to a problem we all face: how
to divide up the limited number of hours in our day between all of the things
we would like to do, or must do to make a living. We simplify the problem by
limiting our objectives to only two things: free time and consumption (similar
to the problem involving Living and Learning in Chapter 3). Because we pay
for our consumption with the wages we receive for working, and working
means not having free time, we face a trade-off: more free time means less
consumption (and of course vice versa).
A trade-off between free time and consumption
Consider a worker, Scott, deciding about how much leisure and consumption
he would like. He is going to accept one job offer from the many job offers he
is fortunate to have. In the jobs open to him, the hours of employment differ
substantially as does the salary. We define h as the fraction of the day that
Scott spends working for wages, with f = 1
h the fraction of that day that
is free time ( f ). Scott consumes his entire income, so his daily consumption,
x, is the total income he would receive if he worked 24 hours, w, times the
faction of the day that he works.
Consumption
x
= Wages ⇥ Proportion of day spent working
= wh
(7.18)
Scott’s utility is given by the following Cobb-Douglas function that expresses
357
Leisure time as a proportion, f = 1 − h
358
MICROECONOMICS
- DRAFT
Figure 7.10: Feasible set and indifference
curves for working time and consumption
The choice of working time (and hence free time)
and consumption. The worker chooses a level of
consumption wh where their consumption is equal
to their working hours (h) multiplied by their wage
(w). The worker balances their consumption (x)
against their leisure time as a proportion of their
day.
f=1
1 − ha
a
mrs = mrt
u3
u2 = ua
u1
xa = wha
x=w
Consumption, x = wh
how he values consumption (x) and free time ( f = 1
u(x, h)
= xa ( 1
h)1
h):
a
(7.19)
Where, as before, 0 < a < 1 and the size of a indicates the relative intensity of preferences for the two goods. Because x = wh we can rewrite this
as
u
= (wh)a (1
h)1
a
(7.20)
In Figure 7.12 we show Scott’s feasible set of choices between consumption
and free time and three indifference curves representing Scott’s preferences
for the two goods. The feasible frontier is your budget constraint. The maximum that Scott could spend on consumption is to have no free time and to set
working time at 1 allowing a total expenditure of w on consumption. So w is
analogous to m in the previous budget constraints and this limits expenditure
on consumption x and the time not working, valued at the wages Scott would
have received had Scott worked the entire day:
Budget constraint:
w
x + (1
h)w
(7.21)
We can re-arrange Equation 7.21 as follows:
Re-arranged constraint
x  wh
(7.22)
Equation 7.22 therefore requires that the value of Scott’s consumption x not
be greater than wages for time worked wh.
M - C H E C K We can rewrite the budget
constraint as
1
h
=
1
x
w
From which we see that
d (1 h)
dx
=
1
w
Which is the effect of greater consumption
on feasible free time.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
The negative of the slope of the budget constraint is the marginal rate of
transformation of reduced consumption into free time. This is how much
additional free time Scott is able to have by giving up one unit of consumption.
The equation for the budget constraint (Equation 7.21) and Figure 7.10 show
that the marginal rate of transformation is as follows:
D (1 h)
= mrt (x, 1
Dx
h)
1
w
=
(7.23)
The slope of Scott’s indifference curves is the marginal rate of substitution
between consumption and free time. As shown in Equation 7.16 the marginal
rate of substitution derived from a Cobb-Douglas utility function is the ratio of
the x exponent to the y exponent times the ratio of the value of the y variable
to the x variable or:
mrs(x, 1
h)
✓
=
◆⇣ ⌘
y
a
x
a
1
The y variable here is the amount of free time 1
h and x is the level of con-
sumption which is x = wh, so
mrs(x, 1
h)
=
✓
a
a
1
◆✓
1
h
x
◆
The best Scott can do in this constrained optimization problem, maximizing
his utility subject to his budget constraint, is to select the bundle (x, 1
h)
such that the marginal rate of substitution is equated to the marginal rate
of transformation. At his utility-maximizing bundle the fraction of the day
Scott will work, h, is equal to a (see M-Note 7.6). As a result, the fraction of
Scott’s day that is free time will be 1
a , because 1
a , the exponent of
"free time" in your utility function is a measure of how important free time is to
you.
M-Note 7.5: Consumption, free time, and work hours
As explained in the previous section, we have:
u(x, h)
Scott’s utility function
wh
Scott’s budget constraint
h)
=
mrs(x, 1
h)
=
using y = 1
MRT (using equation 7.21)
mrt (x, 1
h) =
xa ( 1
x
(1
re-arranged and as an equality
MRS (using equation 7.16)
=
h
=
D (1 h)
Dx
=
h)1
a
x
1
w
✓
◆⇣ ⌘
a
y
1 a
x
✓
◆✓
◆
a
1 h
1 a
x
1
w
359
360
MICROECONOMICS
- DRAFT
Country
Average annual work hours of production workers
Average annual work hours of production workers
Country
France
Germany
Netherlands
Sweden
Switzerland
United Kingdom
3000
2500
2000
1500
Australia
Canada
Japan
United States
3000
2500
2000
1500
1900 1913 1929 1938 1950 1960 1973 1980 1990 2000
1900 1913 1929 1938 1950 1960 1973 1980 1990 2000
Year
Year
(a) Work hours over time in European Countries
To calculate Scott’s time worked, we equate the mrs(x, 1
mrs
Using x = wh
✓
✓
a
a
1
a
1
a
◆✓
◆✓
1
h
x
1 h
wh
wa (1
a
◆
◆
(b) Work hours over time in countries outside of Europe
h) to the mrt (x, 1
=
1
w
=
1
w
h)
=
(1
ah
=
h
a
=
h
h):
Figure 7.11: Work hours over time for a variety
of countries. The data refer to annual average
work hours for full-time production workers
(meaning, excluding supervisory personnel)
Source: Huberman (2004)
mrt
a )wh
ah
(7.24)
Scott’s hours worked only depends on a , or how much Scott values consumption relative
to free time.
Checkpoint 7.5: Preferences for free time and consumption of goods
1.
2.
7.5 Application: Social comparisons, work hours and consumption
as a social activity
In Chapter 3 we looked at data on how men and women spend their time, and
the increase in the amount of time women spend working for pay (the female
labor force participation rate). Here we use the constrained optimization
model to help understand another dramatic change in time use over the last
century.
Figure ?? shows that in every country on which we have data people have
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
361
been working less. But there are important differences among the countries:
• In the Netherlands work hours fell from the equivalent of 62 hours 52
weeks of the year to less than 27 hours per week.
• In Sweden, where work hours also declined dramatically, there was a small
increase in work hours from 1980 to 2000.
• Work hours declined much less in the U.S. than in most other countries –
a decline of 32.77 percent compared to a decline of 58.04 percent in the
Netherlands.
• In the US, as in Sweden there was a slight increase in work hours at the
end of the last century.
How can our model help explain what explains these differences? We modify
the model of choice of work hours to help us understand the differences
among the countries and the changes over time shown in Figure 7.13.
The new idea is that what people consume - the quality, quantity and expense
of what someone wears, or drives, or eats - is a signal to others and to themselves about where they stand in society relative to other people. That is,
people judge their own level of consumption relative to other people’s consumption, not based on the level of consumption alone.
Veblen effects, conspicuous consumption and working time
A particular variant of this view of consumption as social signaling terms the
things we purchase in order to impress ourselves or others conspicuous consumption and it is the high income people who set the standards for everyone
else. One way to model this is to say that we compare our consumption to that
of the very rich, and the closer our consumption is to the consumption of the
rich, the better we feel.
H I S TO RY The term "conspicuous consumption" comes from the American economist
and sociologist Thorsten Veblen (18571929). Well over a century ago he described
exactly the trade off we are modeling here.
"The means of communication and the
mobility of the population now expose the
person to the observation of many persons
who have no other means of judging his
reputability than the display of goods... the
present trend of the development is in the
direction of heightening the utility of conspicuous consumption as compared with
leisure."1
To do this we now define "effective consumption" as how our consumption
feels to us given what others are consuming. To capture this idea, we define
effective consumption as follows:
Effective consumption
x
v
= Consumption
= x
vx
Veblen Effect ⇥ Consumption of the Rich
Where, x is Scott the worker’s consumption, x is the consumption of the rich,
and v is a positive constant representing the Veblen effect.
The negative effect of the consumption of the rich on our utility is captured in
the term v (named after Thorstein Veblen). Effective consumption defined in
this way expresses the idea that the consumption of the rich has the effect of
diminishing the adequacy that we feel for any particular level of consumption.
(7.25)
Leisure time as a proportion, f = 1 − h
362
MICROECONOMICS
- DRAFT
Figure 7.12: Feasible set and indifference
curves for working time and consumption:
Veblen effect. The worker now experiences a
"Veblen effect" which changes their preferences
over work time and leisure, altering the slope of
their indifference curves and therefore resulting in
a new utility-maximizing point. At the new utility
maximum, they consume more goods, work more
and consume less leisure.
f=1
1 − ha
a
b
1 − hv
ua
uv2
uv1
xv = whv
xa = wha
x=w
Consumption, x = wh
This raised the marginal utility of greater consumption to compensate. Scott’s
utility now includes this idea:
u
= (x
vx)a (1
h)1
Using equation 7.16 for the mrs(x, 1
a
= (wh
vx)a (1
h)1
a
(7.26)
h) with a Cobb-Douglas utility function,
Scott’s marginal rate of substitution is now:
mrs(x, 1
h)
=
=
✓
✓
a
a
1
a
1
a
◆✓
◆✓
1
h
◆
xv
◆
1 h
x vx
(7.27)
Because increased consumption by the rich ((x)) has diminished your level
of effective consumption, it has raised the mrs(x, 1
h) (you can see this in
the second term in Equation 7.27). So how much Scott values consumption
relative to how much he values free time is now greater than before. This
means that in Figure 7.12 it has made the indifference curves steeper.
Figure 7.10 shows an initial choice that the worker faces without the Veblen
effect, that is, when he does not worry about what others consume. Figure
7.12, as a contrast, shows how the "Veblen effect" of the consumption of the
rich affects the worker’s choice. The Veblen effect does not alter the feasible
frontier, but it changes the marginal rate of substitution between consumption
and free time. The indifference curves are steeper because at any level of actual (not effective) consumption and free time, the marginal utility of effective
H I S TO RY The people who set consumption
standards, according to Veblen, are the rich.
He wrote: "all canons and reputability and
decency and all standards of consumption are traced back... to the usages and
thoughts of the highest social and pecuniary
class, the wealthy leisure class."2
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
363
consumption has risen. Why? Because when Scott compares himself to the
rich it makes him feel as though he has less effective consumption than he
actually has.
Figure 7.12 shows that when Scott maximizes utility by setting the mrt (x, 1
h) to the mrs(x, 1
h) Scott works longer hours, enjoys less free time, and
increases the amount of consumption he purchases (see the M-Note 7.6).
The time Scott works now is greater than before, as shown in Figure 7.10 (see
the M-Note 7.6):
hv
= a+
(1
a )vx
w
(7.28)
Equation 7.28 shows the working time without the Veblen effect (namely, a )
plus more time that Scott is now motivated to work due to the Veblen effect
(namely, the second term on the right hand side of Equation 7.28).
Did Veblen effects and falling inequality explain declining work hours in the
20th century?
How does this model help us understand how working hours have changed
over time and how work hours differ across countries? The model predicts
that the more that rich people consume, the longer other people will work. So
we would expect people to work longer hours in countries in which the rich
are especially rich, and people to work less where the rich are only modestly
richer than the rest.
Figure 7.13, presents the average annual working hours and a measure of the
fraction of all income received by the richest 1 percent of households in ten
countries over the 20st century. The fraction of income received by the very
rich is a measure of the key variable in the model, the consumption of the very
x
rich divided by the wages of typical workers or w .
The figure shows that this prediction of the model with Veblen effects is borne
out by the data: longer working time is associated with a larger share of income going to the very rich.
But it shows more: the decline in the relative incomes of the very rich is
closely associated with the decline in work hours. Notice that Sweden is
both the most unequal and "longest working" nation (in the early years of the
data set) and also the most equal and country with the most free time (in the
more recent years). The countries that became more equal over this period
also saw the greatest drop in work hours. The increase in work hours in both
Sweden and the U.S. at the end of the last century was associated with an
increase in inequality in both countries.
In this model, conspicuous consumption by the very rich is a kind of "public
bad." It is experienced equally by all members of the society or at least can
F AC T C H E C K In 2001 the tax authorities in
Norway began posting income tax records
online, so that anyone could find out the
income of their neighbors, friends and co
workers. Huge numbers of people accessed
the site. Ricardo Perez-Truglia studied the
statistical relationship between Norwegian’s
income and measures of their "subjective
well-being" – happiness and life satisfaction.
Higher income people were happier and had
greater life satisfaction. But after incomes
became public the differences between rich
and poor in subjective well-being became
much greater.3
364
MICROECONOMICS
- DRAFT
Sweden, 1900
●
Country
Annual Average Work Hours
3000
●
●
Netherlands, 1913
●●
●
●
●
●
2500
●
●
●●
●
●
●
●
2000
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
● ●
●● ● ●
●
●
●
●
● ●
Sweden,
2000
1500
●
●
●
●
●
●
●●
●
● ●
● ●
● ●
●
●●
●
●
●
●
● ●●
●
●
●
●
●
●
●
●
Canada
●
France
●
Germany
●
Japan
●
Netherlands
●
Sweden
●
Switzerland
●
United Kingdom
●
United States
●●
●
●
Australia
●
●
●
●
●
●
●
●
Netherlands,
2000
1.5
2.0
2.5
3.0
Income Inequality
be by anyone with access to TV and social media (it is non-excludable). It is
a "bad" and not a good because it reduces the utility of those it affects (it is
non-rival in the disutility it creates). And the people affected then respond in
ways that generate further negative external effects because the Veblen effect
induces them to work and consume more, increasing the use of our limited
environmental resources.
The Veblen effects model suggests some of the reasons for differing working
hours among countries. But by itself it misses some important parts of the
story. The most important missing element for the decline in work hours in the
20th century is that voting rights were extended to include most adults early
in the century. When overworked employees got the right to vote, in virtually
all countries their trade unions and political parties demanded reductions in
working hours.
The model with Veblen effects and the data provide an illustration of a more
general point: consumption is not just a biological activity. Eating is not just
nutrition. Clothing is not just keeping warm. Your home is more than four walls
to keep out the weather. Consumption is a social activity. As Veblen said our
consumption is a signal to others and to ourselves about who we are. It is
also a social activity in which we engage, for example, for the pleasure of the
company of our friends.
Figure 7.13: Inequality and work hours 1900
to 2000. The hours data are annual average
work hours for full-time production workers. The
income data are based on the share of total income
received by the top 1 percent of households.
Source: Oh, Park, and Bowles (2012).
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
365
M-Note 7.6: The choice of work hours with & without a Veblen effect
Using equation 7.26, Scott’s utility function when deciding between leisure and consumption may be represented as:
u = (x
vx)a (1
h)1
a
With v
0. When v = 0, there is no Veblen effect, and we have same utility function as in
Equation 7.19 that we used in the previous section. When v > 0, there is a Veblen effect.
To calculate the time worked, we need to equate the mrs(x, 1
Following 7.27 and 7.23:
a
mrs
Using x = wh
1
a
1
h) to the mrt (x, 1
(1 h)
a (x vx)
=
1
w
(1 h)
a (wh vx)
=
1
w
h)
=
(1
a )(wh
awh
=
wh
awh
wh
=
aw + (1
hv
=
a+
aw(1
aw
h).
mrt
(1
vx)
(1
a )vx
a )vx
a )vx
w
In the absence of Veblen effect (v = 0), the hours worked h depends only on the importance of consumption relative to free time in the utility function, a . A positive Veblen effect
(v > 0) reduces effective consumption. Because there are diminishing returns to effective
consumption (marginal utility of effective consumption is lower the more of it you have) the
effect of there being less effective consumption is to increase the marginal utility of it. As
a consequence, Scott increases his working hours (and consumption) and reduces his
leisure.
You can also see that the higher the consumption of the rich, the lower is effective consumption, and therefore the higher the hours worked (See Figure 7.12).
Checkpoint 7.6: Veblen effects
• What do you think explains the magnitude of v, the parameter governing the
size of the Veblen effect?
• Juliet Schor, an economist, found that people who watch TV more save less.
Saving is not included in our model, but how do you think this result might be
explained by some kind of Veblen effect?4
• How could our model explain the evidence in the Fact Check about the effects of making incomes public on the relationship between income and
subjective well being? How would that affect v the Veblen effect coefficient?
7.6 Quasi-linear utility and demand
?? We don’t always want to know how people trade off the benefits of two
commodities – like coffee and data – in their utility functions. Sometimes,
we find it useful to think about how a person will trade off money left over for
other purchases (y) and a commodity (x), like we saw earlier in the example of
kilograms of fish (x) and money for other goods (y). We explore this idea using
Q UASI - LINEAR FUNCTION When a function
is quasi-linear it depends linearly on one
variable, e.g. y, and non-linearly on another
variable, e.g. x, and has the form u(x, y) =
y + g(x). Hence it is quasi or "partly" linear.
366
MICROECONOMICS
- DRAFT
Figure 7.14: Harriet’s indifference curves: quasilinear utility. With quasi-linear utility, marginal
rates of substitution depend only on the amount
of the good x, and not at all on the amount of
money left over to buy other goods, y. As a result
indifference curves with different levels of utility
are vertical displacements of a single curve. This
means that the slopes of the indifference curves
when consuming x0 amount of fish are the same
independently of the amount of money she has for
other purchases.
y3 = 17
Money left over, y
y2 = 15
y1 = 13
u3
u2
u1
x0
x1
Kilograms of fish, x
a Quasi-linear (QL) utility functions with the form:
u(x, y) = y + g(x)
(7.29)
Utility in the quasi-linear function depends linearly on y. The more y Harriet
has, the higher her utility. Her marginal utility for y is always 1 and it does not
decline as she gets more y. These properties make y more like wealth measured in terms of money than like a particular good such as data or coffee, so
we will often refer to y as money, understanding that it is really generalized
purchasing power that can be spent on many other things possibly in many
periods.
The analysis of demand is greatly simplified if the marginal rate of substitution
depends only on the amount of the good someone purchases, and not on the
amount of money she has left over. Here are the simplifications:
• Prices: When y is money left over for other purchases, then py = 1 (the
p
price of a dollar is a dollar), so we can simplify mrt (x, y) = px = px = p.
y
That is, when the other good is money, we shall simply refer to the price of
good x as p, which is the opportunity cost of x.
• Marginal utility : when utility is quasi-linear in y, then uy = 1, that is, the
marginal utility of money is constant and equal to 1. Therefore the marginal
rate of substitution mrs(x, y) = uux = ux is just the marginal utility of x, which
y
is the trade-off of money for x.
Q UADRATIC, QUASI - LINEAR UTILITY In
the case of quadratic, quasi-linear utility,
the non-linear part of the utility function,
g(x) is a quadratic function of x such
as g(x) = p̄x 12 ( p̄/x̄) x2 , where x̄ is
the maximum amount of x someone is
willing to consume and p̄ is their maximum
willingness to pay for x when they have
not yet consumed any x. The quadratic
quasi-linear utility is quadratic in x (it has a
"squared" term in x) and linear in y.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
367
• Willingness to pay : Because with the quasi linear utility function the mrs
is simply ux (x, y), that is, the marginal utility of x when consuming the
bundle (x, y), the mrs is also the maximum amount (in money units) that
the person would be willing to pay to have a small increase in the amount
of x.
If we use a quasi-linear utility function in which money left over is the linear
good, then utility is measured in money. To see this suppose the person
spent nothing on the other good; then Equation 7.29 tells us that u = y. Utility
is the amount of money the person has to spend. It is also the case that if
the amount of money the person has to spend increases by $10, then utility
increases by 10. So the units in which utility is measured is money.
This does not mean that the only thing the person cares about is money.
Whatever x is may matter a lot. It is just that how much it matters will be
measured in money equivalents.
Quadratic, quasi-linear utility
p̄, x̄, SATIATION & BLISS
Many of the examples in this book use the particular class of quadratic quasilinear utilities, where the function g(x) = p̄x
1
2
p̄
x̄
x2 is a quadratic function
of the good x, and as a result the utility function is:
Quadratic, quasi-linear utility
u(x, y) = y + p̄x
pay for good x. She won’t pay more than
p̄ to get a unit of x.
x̄ is the person’s satiation point for x,
✓ ◆
1 p̄ 2
x
2 x̄
(7.30)
In Equation 7.30:
• Satiation: The parameter x̄ represents the level of x at which the buyer is
satiated with x and would consume no more even if the price were zero.
• Maximum willingness to pay: The parameter p̄ > 0 represents the
buyer’s maximum willingness to pay for the first unit of x when they do not
have any x, i.e. when x = 0.
The marginal utility of x with the QQL utility is:
Du
= ux (x, y) = p̄
Dx
p̄ is the person’s maximum willingness to
p̄
x
x̄
(7.31)
Equation 7.31 tells us the following:
• When x < x̄, the buyer’s marginal utility of x is positive, and she regards x
as a good.
• When x > x̄, the buyer’s marginal utility of x is negative, and she regards x
as a bad.
If y is budget left over to buy other goods, then the marginal utility of y is
always 1, regardless of the levels of x and y. As a result, the marginal rate of
beyond which her marginal utility of x
is negative. She would prefer not to
consume x > x̄.
The point at which you are sated (verb)
is where you reach satiation (noun) from
consuming a good, like x. The intuition is
easily seen with food: you reach satiation
at that point where you do not want to eat
another mouthful (the marginal utility hits
zero) or, if you do, you know you’ll regret it
(the marginal utility will be negative). Or, it is
the point at which you have reached bliss,
which is perfect happiness or great joy, and
at which, if you consumed or did any more, it
would detract from that bliss, joy or wonder.
368
MICROECONOMICS
p
- DRAFT
Figure 7.15: Harriet’s marginal rate of substitution (demand): quadratic quasi-linear
preferences. The figure shows Harriet’s mrs(x, y)
for a good, x, and money for other goods, y. That
is, the downward-sloping line is Harriet’s willingness to pay in money for an additional unit of good
x for different levels of the quantity she has of
good x changes and is therefore also her demand
curve because it shows the relationship between
the price of the good and how much of it she will
buy at different prices. The vertical intercept, p̄,
is the person’s maximum willingness to pay when
they currently consume zero units of good x. The
horizontal intercept, x̄, is the person’s satiation
point or bliss point beyond which Harriet’s marginal
rate of substitution is negative so that x changes
from being a good to a bad.
Price per unit of good x, p
p = Maximum
Willingness to Pay
⎛p⎞
mrs(x, y) = p − ⎜ ⎟x
⎝x⎠
x = Point
of Satiation
Quantity of the good, x
x
substitution is equal to the marginal utility of x:
mrs(x, y) =
ux
= ux = p̄
uy
p̄
x
x̄
(7.32)
We can think about Equation 7.32 in the following way:
• Equation 7.32 is the equation for a line.
• Equation 7.32 has vertical intercept p̄, which is the buyer’s maximum
willingness to pay.
• Equation 7.32 has a horizontal intercept x̄, which is the point beyond which
the buyer does not want to pay for good x (at x = x̄, the buyer’s willingness
to pay is zero). x̄ is the buyer’s bliss point. To get the buyer to consume
more than x̄ of the good, you would have to pay her, rather than expecting
her to pay you.
• Equation 7.32 shows that the marginal rate of substitution has a slope of
p̄
x̄ ,
which is the negative of the ratio of the maximum willingness to pay to
the satiation point.
Equation 7.30 shows that Harriet’s utility for the good x depends on how much
she has relative to a target level x̄. When 0  x  x̄, Harriet’s utility increases
if she has more x. But when x
x̄, Harriet’s utility decreases as she gets
more x. We can plot quadratic, quasi-linear marginal rate of substitution or
willingness-to-pay as a function of x as in Figure 7.15.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
369
The inverse demand function, which gives the highest price Harriet will pay for
each given total amount of the good is:
p(x, m) = p̄
Inverse demand function:
p̄
x
x̄
(7.33)
We will often simplify the inverse demand function such that the slope of the
p̄
function, x̄ is represented by a simple slope coefficient, b . We will therefore
represent the demand curve as follows:
Demand curve:
p(x, m) = p̄
bx
(7.34)
Suppose Harriet has quadratic quasi-linear preferences between a good x
and money left over to buy other things, with the willingness-to-pay p(x) =
mrs(x, y) = p̄
p̄
x̄ x.
She starts out with budget m and has the opportunity
to buy any amount of the good x at the price p. If p(0) = p̄ > p, Harriet
will buy at least some of the good. If she buys x0 units of the good, Harriet’s
willingness-to-pay will have fallen to p(x0 ). If p(x0 ) > p, then Harriet will buy
more of the good, and so on, until p(x) = p.
At this point Harriet doesn’t buy any more of the good, and spends the rest of
her budget on other things. So from Harriet’s willingness-to-pay (or marginal
rate of substitution), we can derive the quantity demanded at a market price
p, which tells us how much of the good she will buy when the market price for
the good is p.
x( p) = x̄
x̄
p
p̄
(7.35)
We can also re-write Equation 7.33 as a demand function
Demand function
x(m, p)
= x̄
x̄
p
p̄
(7.36)
Equation 7.36 says that for p > 0 a person will consume an amount less than
their point of satiation (x̄) given the price of the good p and their maximum
willingness to pay p̄. As the price of x increases, then the quantity demanded
of x will decrease.
Figure 7.16 demonstrates this relationship by showing Harriet’s utility maximizing choices between x and y with her indifference curves and budget constraints for three prices of x in the top panel, while also showing her marginal
rate of substitution of money for the good in the lower panel. The lower panel
also shows how Harriet’s marginal rate of substitution corresponds to a demand curve, by showing three different price levels and how the given price
determines the quantity demanded at that price.
M - C H E C K For example, when the market
price is: p = $10, p̄ = $20, and x̄ = 10,
then Harriet would like to buy 5 apples
since her willingness-to-pay is the following:
mrs(5, y) = $10 for any y.
Money left over, y
370
- DRAFT
MICROECONOMICS
Figure 7.16: Harriet’s utility-maximizing choice
and marginal rate of substitution (demand):
quasi-linear preferences. The top figure shows
Harriet’s indifference curves for kilograms of fish
(x) and money left over for other goods (y). The
figure shows her utility-maximizing choices at three
levels of prices for a kilogram of fish. Harriet’s
2
utility function is u(x, y) = y + 20x 12 20
10 x .
Her budget constraint is y = 600 px x. The lower
panel shows Harriet’s mrs(x, y) for a good, x, and
money for other goods, y. Harriet’s marginal rate
of substitution is therefore mrs(x, y) = 20 2x.
Her marginal rate of substitution, as her willingness
to pay in money (y) for goods (x) is her demand
function for x. She has a y-intercept of y = 20 = p̄
(her maximum willingness to pay) and her xintercept is x = x̄ = 10 (the amount of fish that
satiates her appetite for fish, which is also the
maximum quantity of fish she would consume were
the price of fish zero). The slope of her marginal
rate of substitution suggests she will exchange
money (y) for fish (x) until her mrs(x, y) = px , i.e.
when 20 2x = 10, which implies x = 5 when
px = 10.
600
u3
558
a
c
550
b
u2
bc3
pL = 6
u1
bc2
pB = 10
bc1
pH = 14
xH = 3
Price per unit of the good, p
p = 20
pH
xB = 5
xL = 7
x = 10
mrs(x, y) = 20 − 2x
High price, pH = 14
f
Base price, pB = 10
e
pB
g
pL
xH = 3
xB = 5
Low price, pL = 6
xL = 7
Quantity of the good, x
x = 10
M-Note 7.7: The demand for x and y with with quadratic quasi-linear preferences
We begin with the mrs = mrt rule, and then rearrange it to give us the demand for good x
mrs(x, y)
=
p̄
p̄
x
x̄
p̄
x = p = mrt (x, y)
x̄
=
p̄
p
x(m, p)
=
x̄
x̄
p
p̄
(7.37)
The person will then use whatever remains of their budget (m) as money to spend on
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
371
other goods, y, given what they spent on x at its price, p:
y
=
m
y
=
m
y(m, p)
=
m
px
✓
p x̄
◆
x̄
p
p̄
2 x̄
px̄ + p
p̄
(7.38)
Equation 7.38 shows that once we determined the demand for x, we can derive the
demand for the other goods as well.
7.7 Price changes: income and substitution effects
When the price of a good changes, the consumption of the goods changes as
we saw when deriving the demand curve from the offer curve in Figure 3.13.
The total amount of the change is made up of two components:
• The income effect
• The substitution effect
The income effect occurs because a person has more or less purchasing
power as a consequence of the change in prices. A person’s real budget is
a measure of their purchasing power with a given budget – the maximum
amount they can purchase of a good at given prices. If the price of the good
I NCOME AND SUBSTITUTION EFFECTSWhen
the price of a good changes there are two
effects. The first, due to the change in
relative prices (the good’s price relative
to other goods’ prices), leads people to
substitute among the goods they buy. This
is the substitution effect. The second effect
arises because the price change alters
peoples’ real income, expanding or shrinking
the feasible set of purchases. This is the
income effect. The two effects are calculated
so that they sum to the total change in the
quantity of the good consumed.
increases the consumer can buy less of the good than before the price increase. They have less purchasing power at their fixed budget. Generalizing
this to two goods, if the price of one good decreases while the price of the
other good remains the same, the person can buy more of both goods when
the price of one good decreases.
The substitution effect occurs when, as the prices of goods change, people
substitute away from goods that are relatively expensive and they will substitute towards goods that are relatively cheap. So, when choosing between two
goods that both provide positive utility, when the price of one increases relative to the other good, the person will normally substitute away from that good
toward the other good (assuming some degree of substitutability between the
goods).
Income and substitution effects for normal goods
Consider a change in the price of good x, for example, an increase in px . Harriet will change the amount of good x that she buys. We can decompose or
separate this change into a substitution effect, which represents how Harriet
would change her choice if she had to move along the original indifference
curve, and an income effect, which represents the change in her behavior due
to moving to another (in this case, lower) indifference curve. Harriet starts at
point a with bundle (xa , ba ) on her initial budget constraint bca before the price
C OMPENSATED BUDGET CONSTRAINT The
compensated budget constraint takes the
new prices of goods as given (it is parallel to
the budget constraint after the price change),
but it gives the person sufficient income to
return to their original indifference curve,
therefore creating a new point of tangency
with the original indifference curve. In the
case of a price decrease a compensated
budget constraint would take money away
from a person, for example via a lump sum
tax.
372
MICROECONOMICS
- DRAFT
Gigabytes of data, y
u1
c
yc
yb = ya
Figure 7.17: Income and substitution effects:
Cobb-Douglass utility. The substitution effect
is shown by the hypothetical movement along u2
from a to c. The income effect is shown by the
movement to another indifference curve, from b
to c with the income effect being the difference in
the purchases of coffee between those two points.
The fact that the change in the price of coffee does
not affect the level of purchases of data ya = yb
is an attribute of the specific Cobb-Douglas utility
function we have used here, it is not a general
result.
u2
a
b
bca
initial
budget
bcb
budget
when px
increases
xb
xc
Income
effect
bcc compensated
budget
xa
Substitution
effect
Kilograms of coffee, x
increases. When the price increases, the budget constraint pivots inwards
to bcb and creates a new equilibrium at b with bundle (xb , yb ). The income
effect can change both the amount of good x and good y, as Figure 7.17
shows.
To decompose this change into an income and a substitution effect, we need
to construct a counter-factual. We take Harriet back to her original indifference curve, but we retain the slope of the new budget constraint to see how
the new prices would have affected her. To break down the effects, we use
the idea of a compensated budget constraint. The compensated budget
constraint could hypothetically be implemented by a policy-maker who wants
to ensure that a person who has lost real value of their available budget because of price increases can be compensated in some way, for example, by a
government transfer.
The compensated budget constraint bcc allows us to consider how Harriet
would respond to the new price for good x if she were limited by the compensated budget constraint that provides Harriet with just enough money to be on
her original indifference curve, u2 . The new tangency is at point c at bundle
(xc , yc ).
The difference between xa and xb is the total effect of the price change or
the substitution effect plus the income effect. By construction the substitution
effect causes a movement along the original indifference curve to point c. The
E FFECTS OF A PRICE CHANGE The total
effect of a price change is the change in
quantity demanded. The decomposed
effects shows how the total effect is broken
up (decomposed) into the two parts of the
substitution effect (a movement along the
indifference curve) and the income effect (a
movement to a new indifference curve).
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
Utility Function
Income Effect
Substitution Effect
Cobb-Douglas
Yes
Yes
Quadratic, quasi-linear
No
Yes
Perfect Complements
Yes
No
373
Table 7.1: Utility Functions and their income
and substitutions effects. Remember that the
substitution effect is captured by a movement along
an indifference curve as prices or real incomes
change, whereas the income effect is captured by a
movement to a new indifference curve.
difference between xc and xa is the substitution effect. The difference between
xb and xc is the income effect, which is what drives Harriet’s choice to point b
with bundle (xb , yb ).
<Insert links to online appendices with no income effect and no substitution
effect>.
Complements and substitutes in consumption
The size of the income and substitution effect will depend on whether the
goods "go together" or have an "either/or" quality. If two goods are more
enjoyable consumed together, then they are complements (coffee and cookies). "Either/or" goods are called substitutes: they are consumed instead of
R E M I N D E R In Chapter 6 we introduced
the idea that in inputs to a production
process may be either substitutes (computer
driven machine tools and skilled machine
operators) or complements (computer driven
machine tools and engineer-programmers).
each other (tea and coffee). At the extreme, perfect substitutes have a constant marginal rate of substitution (linear indifference curve), whereas perfect
complements (right shoes and left shoes) are only valuable when consumed
together. Indifference curves for perfect complements are L-shaped.
M-Note 7.8: Complements and substitutes in consumption
Complements: Cookies and coffee. In the Cobb-Douglas utility function u
(coffee) and y (tea) are complements because
∂u
= axa
∂x
=
xa y1
a,
x
1 1 a
y
which, because 0 < a < 1, means that
∂ u/∂ x
= (1
∂y
a )axa
1
y
a
= (1
x
a )a ( )a > 0
y
so the greater is the consumption of y the higher is the marginal utility of x. By the same
reasoning, the greater is the consumption x, the higher is the marginal utility of y.
Substitutes: Coffee and tea. Here is a utility function in which x (coffee) and y (cookies)
are substitutes:
u = (x + ey)a
where e is a positive constant measuring how much the person prefers tea to coffee and
0 < a < 1. Therefore, finding the marginal utility of x by taking partial derivatives:
∂u
= a (x + ey)a
∂x
1
>0
The marginal utility of x is positive. But how does the marginal utility of x change as
consumption of y changes? We can work that out by taking the partial derivative of the
marginal utility of x with respect to y:
∂ u/∂ x
= e (a
∂y
1)a (x + ey)a
2
<0
It is negative because a < 1. This shows that the marginal utility of coffee is less the more
tea the person consumes. The same reasoning shows that the marginal utility of tea is
C OMPLEMENTS AND SUBSTITUTES IN
CONSUMPTION Goods are complements in
consumption if an increase in the quantity
consumed of one raises the marginal utility
of the other. Goods are substitutes in
consumption of an increase in the quantity
consumed of one reduces the marginal
utility of the other. The property of being
complements or substitutes is symmetrical:
If good x is a complement of good y, then y
is also a complement of good x. The same is
true for substitutes.
374
MICROECONOMICS
- DRAFT
less, the more coffee the person consumes. For this particular utility function, tea and
coffee are what is called perfect substitutes. In this case,
∂u
= ea (x + ey)a
∂y
1
so the individual’s marginal rate of substitution, the ratio of the two marginal utilities, is as
follows:
∂ u/∂ x
a (x + ey)a
=
∂ u/∂ y
ea (x + ey)a
1
1
=
1
= mrs
e
Because the mrs(x, y) is the negative of the slope of an indifference curve, the fact that it
is constant means that the indifference curves are linear. This is what the fact that x and y
are perfect substitutes means.
7.8 Application: Income and substitution effects of a carbon tax and
citizen dividend
The decomposition of the results of price changes into income and substitution effects can be illustrated by the proposed carbon tax to reduce emissions
of carbon dioxide and other greenhouse gases that contribute to climate
change.
The prices of petroleum, coal, natural gas and other fossil fuels do not include
the costs of the environmental and climate-change external effects of their use
. This means that people pay a private cost of using fossil fuels that is lower
than the social (private plus external ) cost of using them. The result – as in
the case of over-fishing in Chapters 1 and 5 – is overuse of fossil fuels.
We now consider tax imposed on the sale of fossil fuels, in conjunction with a
transfer of the tax resulting revenues in equal amounts back to the members
of the population, called a citizen’s dividend. We ask: how would this so-called
carbon tax and citizen dividend policy reduce consumption of fossil fuels and
affect citizens’ consumption of other goods.
We consider two steps in our policy process:
• The substitution and income effects of the increased price of fossil fuels.
• The income effect of the citizen dividend.
Reducing carbon emissions by imposing the tax
In Figure 7.19, the fossil fuel consumption (x) of a citizen is plotted on the
horizontal axis, and the consumption of others goods measured in some
currency is plotted on the vertical axis (y). Before the tax, a citizen is at point a
in Figure 7.19 where the marginal rate of substitution equals the marginal rate
of transformation, which is the existing price of fossil fuels, pa . At a, the citizen
has a utility of u2 on the corresponding indifference curve.
The government then imposes a tax on fossil fuels. The tax increases the
Figure 7.18: A conservative case for climate
action. In their 2017 op-ed (opinion editorial) in
The New York Times three economists made "A
conservative case for climate action," proposing
the carbon tax and citizen dividend that we analyse
here. They included Martin Feldstein and Gregory
Mankiw, chief economic advisors to U.S. Presidents
Ronald Reagan and George W. Bush, respectively.
R E M I N D E R In Chapter 5 we explained that
the social cost equals the private cost plus
the (negative) external cost imposed by a
person’s action. In that case, a person’s
marginal private cost of fishing included
their disutility of fishing, but the social cost
included not only the private costs but also
the negative external effect the fishermen
imposed on others.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
md
mb = ma
c
yc
yb = ya
a
b
bcb
budget
when px
increases
bcc
compensated
budget
bca
initial
budget
u1
mb
d
yd
b
yb
xb
effect
Fossil fuels consumed, x
(a) Substitution and income effects of the price increase
a
bcd
budget with
dividend
bcb
budget
with tax
increase
xb Income xcSubstitutionxa
effect
u2 u3
dividend
u2
Consumption of other goods, y
Consumption of other goods, y
u1
mc
375
xd
Fossil fuels consumed, x
(b) Second income effect due to citizen dividend
price of fossil fuels. With the increase in the price from pa to pb due to the the
tax, the citizen’s budget constraint becomes steeper. It pivots inward round
its y-axis intercept because the amount of the budget itself is unaffected, but
the price has increased. So bcb is the new budget constraint. As before, citizen now maximizes her utility now consuming at point b where her marginal
rate of substitution equals the new marginal rate of transformation, pb . At b
the constrained utility maximum, the citizen has decreased her consumption of fossil fuels from xa to xb , consistent with the policy goals of the tax to
decrease consumption of fossil fuels.
Setting aside the value that the citizen places on the overall mitigation of
the the climate change crisis that the population wide effects of the policy
accomplished, the policy has lowered her utility because she is on a lower
indifference curve, u1 < u2 .
Is it fair?
This will be true of all citizens, but the effect will differ across by levels of
income. In the U.S., the reduction in real income imposed by the carbon tax
on will be a larger among poorer households. This is because, as Figure 7.20
shows:
• while (panel a) higher income people spend much more than lower income households on carbon costs (think about air travel, heating and air
conditioning large houses)
• ...expenditure on carbon costs as a fraction of their total expenditure is
greater for lower income households (panel b).
As a result high income people will pay more of the tax than low income
people, but the tax will lower the real income of poor people by a larger per-
Figure 7.19: Carbon tax with dividend. A
citizen decides on consumption of fossil fuels
(x) other goods (y). Prior to the introduction of
the carbon tax the citizen’s budget is ma and the
budget constraint is the line labeled bca , with
slope pa . The utility-maximizing allocation is
on the budget constraint at point a where the
citizen’s indifference curve u2 is tangent to the
budget constraint. A carbon tax is proposed that
increases the price from pa to pb steepening the
budget constraint. The result, shown in panel
a, is the budget constraint pivots inwards. The
effect of the price change – a reduction in fossil
fuel consumption from xa to xb – is the sum of the
income and substitution effects. In panel b the
citizens’ dividend increases the household’s budget
so the budget constraint shifts outwards and has a
vertical intercept (the budget itself) md > mb . The
indifference curves shown here are based on a
Cobb-Douglas utility function with a = 0.5.
376
MICROECONOMICS
- DRAFT
centage that will be the case for higher income people. In the U.S. the carbon
tax is regressive, meaning that the amount paid as a fraction of a household’s
income is greater for lower income households.
This is where the citizen’s dividend comes in.
Increasing income and ensuring fairness through a citizen’s dividend
To see how the citizens’ dividend alters the result, turn to Figure 7.19 b. As
in panel a, the citizen is at point b with lower utility than before the tax. But
now suppose the total carbon tax revenues collected are divided equally and
distributed equally to each household as in the amount mb to md . This is the
citizen’s dividend. As you can see the effect is to shift upwards the budget
constraint by the same amount.
With the higher income, the citizen maximizes her utility at point d. For the citizen we have modeled, the dividend mb to md is large relative to her previous
budget. The result is an increase in consumption of other goods, so that her
level of utility is higher than it was before the tax, that is, comparing points d
and a, u3 > u2 . At d, they have higher consumption of other goods (yd > yb )
and they consume the a lower level of fossil fuels than they did previously,
but greater than before the dividend (xa > xd > xb ). With the greater consumption of other goods, the citizen has higher utility and they obtain utility of
u3 > u1 .
All citizens will receive the same dividend. This means that more than half of
the households – those poorer than the mean income – will experience an
increase in their disposable income (income after paying taxes and receiving
the dividend) as a result of the carbon tax and citizens’ dividend policy. They
will pay less in taxes (which are proportional to the cost of the carbon they
consume) than they receive in the citizen’s dividend which is proportional to
the mean carbon costs consumed in the population and hence equal for all
citizens.
While the carbon tax alone is regressive,the carbon tax and the citizen’s dividend taken together is progressive. The case we modeled in Figure 7.19
illustrates the increase in disposable income and utility of a poorer than average citizen.
International differences
In some countries, however, the picture is reversed, with poor people spending a small fraction of their budget carbon related consumption and wealthy
families spending a larger share on fuel as a relative proportion of their income. The data in Figure 7.21 for Mexico show exactly this, at least for motor
fuels. As with carbon costs as a whole, the U.S. data (in the left panels) show
F AC T C H E C K Using detailed data on
household expenditures economists Anders
Fremstad and Mark Paul calculated that a $
50 per ton of CO2 would leave 56 percent of
all U.S households with a higher real income
(the same would be true of 84 percent of
households with incomes less than the
mean.)5
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
377
15%
Cost as percent of income
Cost in $ per person
4000
3000
Mean
2000
Median
1000
0
10%
5%
0%
1
2
3
4
5
6
7
8
9
10
1
2
Decile
(a) Average carbon costs per person by income decile
3
4
5
6
7
8
9
10
Decile
(b) Average proportion of expenditure on carbon by income decile
that the fraction of the household budget spent on electricity and motor fuels respectively falls dramatically as income rises. But this is not the case
for Mexico. The fraction of the household budget spent on electricity is only
modestly lower for the upper income households. And for motor fuels (car and
truck use) the fraction of the budget spent is much greater for high income
households than for those with lower incomes (among whom car and truck
ownership is limited).
Figure 7.20: Dividend distribution and carbon
costs. The left-hand figure shows the absolute
amount spent on carbon for each of the ten
household income deciles (1 is poorest, 10 is
richest). The right-hand figure shows the proportion
of household expenditure on carbon by for the ten
income deciles. Source: Fremstad and Paul (2019)
using data from the Energy Information Agency, the
Bureau of Economic Analysis, and the Bureau of
Labor Statistics.
7.9 Application: Giffen Goods and The Law of Demand
The demand curves you have seen all slope downward: a lower price is associated with more purchases. This is called the Law of Demand, and the
movement of prices and quantities purchased in opposite directions that it predicts is widely observed. But there is a special kind of good – called a Giffen
good – for which the law of demand is violated. For Giffen goods, a higher
price is associated with a greater amount of purchases.
You already know that for an inferior good the amount purchased will decline
as income rises. This is not really surprising, some of the low cost foods that
people eat when they have very limited budgets will not be purchased at all
when they have more income to spend. A Giffen good really is surprising
because less is purchased when its own price decreases.
How could this be?
Think about a poor family consuming a large amount of some inferior good.
When the price decreases, as you know, there is both a price effect motivating
the family to purchase more and an income effect resulting from the decrease
in price. Because the good is inferior, the higher real income of the family
motivates them to purchase less of the good. If the negative income effect is
L AW OF D EMAND The law of demand holds
that a decrease in the price of a good will
be result in an increase in the quantity of
the good purchased. Giffen goods are an
exception to the law.
378
MICROECONOMICS
- DRAFT
Average Electricity Expenditure (% of Total)
Average Electricity Expenditure (% of Total)
2.0
6
4
2
0
1.5
1.0
0.5
0.0
1
2
3
4
5
6
7
8
9
10
1
2
3
Expenditure decile
5
6
7
8
9
10
(b) Mexican Consumption of Electricity
Average Motor Fuels Expenditure (% of Total)
(a) U.S. Consumption of Electricity
Average Motor Fuels Expenditure (% of Total)
4
Expenditure decile
6
4
2
0
4
3
2
1
0
1
2
3
4
5
6
7
8
9
10
1
2
3
Expenditure decile
(c) U.S. Consumption of Motor Fuel
4
5
6
7
8
9
10
Expenditure decile
(d) Mexican Consumption of Motor Fuel
greater than the positive substitution effect, purchases will decline in response
to a decline in the price.
Here is an example. In China, for very poor households, rice is the main
staple food and if they have enough money, they add other foods that make
rice taste better, such as shrimp or beef. However, when the price of rice
increases, this means the households have little money left over to buy beef or
shrimp. Consequently, they will consume more rice even though it’s price has
increased. As a result, over some range of prices the demand curve for rice
for these families is upward-sloping. Of course if the price of rice rose so high
that the household purchased only rice, then further price increases would
have to reduce the amount purchased, so the demand curve would then be
downward sloping as the Law of Demand requires. Just such a demand curve
Figure 7.21: U.S. and Mexican consumption
of motor fuel. Each county’s income distribution
is divided into deciles from poorest (1) to richest
(10). The average consumption of motor fuel as
a proportion of total family income for each decile
is shown by the size of the bar for that decile. In
the U.S., the consumption as a share of income
is higher for lower deciles than for higher deciles.
In Mexico, the consumption as a share of income
is lower for lower deciles than for higher deciles.
Source: Pizer and Sexton (2019) using data from
the
U.S.
Expenditure
REM
I N DConsumer
E R An inferior
good Survey
is one (Bureau
for which
of
Labor Statistics)
Mexico’s
National
Survey
purchases
fall asand
income
rises.
A Giffen
of Income and Expenditure (Instituto Nacional de
good
must
be
inferior,
but
not
all
inferior
Estadistica y Geografia).
goods are Giffen goods.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
is illustrated in Figure 7.22.
This is exactly what economists who studied subsidies of rice observed in
Hunan, a region of China. They ran an experiment by subsidizing the price
of rice, lowering the price the families actually paid. When they provided the
subsidy, very poor households reduced their consumption of rice. That is,
379
G IFFEN GOODOver some range of prices,
purchases of a Giffen good increase if
the price rises, and fall if the price falls.
Giffen goods are an exception to the law of
demand.
p
the rice subsidy so that prices rose, the households consumed more rice. For
these households rice was a Giffen good.6
7.10
Market demand and price elasticity
at that price of all the people making up the demand side of the market. We
can compute the market demand by adding up the individual demand curves.
If we plot all the demand curves with the quantity demanded of x on the horizontal axis and the price p on the vertical axis, this requires the horizontal
summation of the individual demand curves. We use an upper-case X for
market demand and a lower-case x for an individual demand.
Figure 7.23 shows how (on the left) an individual market demand curve is (on
the right) summed over ten people to produce the market demand curve, that
is X ( p) = x1 ( p) + x2 ( p) . . . + x10 ( p).
A Linear Market Demand Curve (Quadratic quasi-linear utility)
If there are n identical buyers, each of whom has the same quadratic, quasilinear utility for the good, with the same parameters x̄, p̄, each individual has
the demand (from Equation 7.32):
x(m, p) = x̄
x̄
p
p̄
(7.39)
The market demand is then the sum of all the individual demands. But, since
they are all equal for the identical people, this is the same as the the number of people (n) multiplied by the individual demand curves. In the quadratic
quasi-linear case, therefore, the market demand curve is a given by the following:
Market demand
X ( p, m, n)
= Number of people ⇥ Individual demands
✓
◆
x̄
= n x̄
p
p̄
X̄
= X̄
p
(7.40)
p̄
Because the market demand is the summation of individual demands, the
market demand function is also downward-sloping: quantity demanded falls as
the market price increases.
Giffen demand
Downward−
sloping
demand
Upward−
sloping
demand
The market demand for a good at any given price is the sum of the demands
Individual demand
Price per unit, $, p
when rice was cheap, households consumed less of it. When they removed
x
Quantity of the good, x
Figure 7.22: Demand for a Giffen good. The
demand for a Giffen good is upward-sloping
when the price is low. Over this region, a higher
price is associated with more purchases. At a
sufficiently high price, though, demand becomes
downward-sloping.
M ARKET DEMAND CURVE The market
demand curve is the horizontal summation
of individual buyers’ demand curves. That is,
for each price (on the vertical axis) we add
together each person’s quantity demanded
at that price
380
MICROECONOMICS
- DRAFT
p = 10
p = 20
Price per kilogram, p
Price per kilogram, p
p = 20
Market price, p = 10
i
●
1
Demand: x(p) = 10 − p
2
Inverse Demand: p(x) = 20 − 2x
5 10
Demand, X(p) = 100 − 5p
1
Inverse demand, p(X) = 20 − X
5
p = 10
●
(b) Market demand
Re-arranging Equation 7.40, we can find the inverse market demand function:
p(X )
=
p̄
p̄
X
X̄
(7.41)
The inverse market demand curve is linear with a vertical intercept of p̄ (the
maximum willingness to pay of buyers like Harriet), a horizontal intercept of X̄ ,
Dp
p̄
.
X̄
M-Note 7.9: Market demand with 10 buyers
Let us assume that the fish market is made up of Harriet and 9 other buyers who are
identical to her (a total of ten buyers). Harriet’s quadratic, quasi-linear demand function
was:
Harriet’s Demand:
x( p) = 10
1
p
2
(7.42)
If all the fish buyers are identical to Harriet, then we can sum their demand functions
(quantity as a function of price), xi ( p) to get the market demand, X ( p). This is the same
as multiplying the demand function by the number of people, n = 10, to get the market
demand function:
X ( p)
=
=
=
=
X = nx = 100
Kilograms of fish, X
(a) Individual demand
and a slope of Dx =
Market price, p = 10
X* = 50
Kilograms of fish, x
Inverse market demand
i
n(xi ( p))
1
n(10
p)
2
1
10 ⇥ (10
p)
2
100 5p
(7.43)
Recall, though, that we typically graph price as a function of quantity, or the inverse demand function. We use the market demand curve to find the inverse market demand curve
with price as a function of quantity by re-arranging Equation 7.43 and similarly for Harriet
Figure 7.23: Individual and market demand. In
figure a., we present Harriet’s demand at different
prices per kilogram of fish. On the right, is the
market demand for fish, which is the sum of ten
identical fish-buyers’ demands for fish (including
Harriet). Notice that Harriet’s individual demand
curve is much steeper than the market demand
curve. The change occurs because, for example,
for every $2 decrease in the price Harriet will buy
one more unit of fish, for the market as a whole,
each of 10 people would buy one more unit of the
good.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
1
p(X) = 20 − X
5
p = 20
Price per unit of the good, p
p
Price per unit of the good, p
Elastic
η >1
Unit Elastic
η =1
●
Inelastic
η <1
pf = 14
●
⎛6⎞
ηg = 5⎜ ⎟ = 0.43
⎝70⎠
e
pg = 6
●
x
Quantity of the good, X
⎛10⎞
ηe = 5⎜ ⎟ = 1
⎝50⎠
●
Xf = 30
0
⎛14⎞
ηf = 5⎜ ⎟ = 2.33
⎝30⎠
f
pe = 10
381
g
Xg = 70
Xe = 50
X = 100
Market quantity of the good, X
(a) General values of price elasticity
(b) Price elasticity for 10 People like Harriet in a market
with re-arranging Equation 7.42:
Harriet’s inverse demand:
p(x)
=
20
Market inverse demand:
p(X )
=
20
2x
1
X
5
(7.44)
(7.45)
Contrasting Equations 7.44 and 7.45 we can see that they have identical vertical intercepts equal to p̄, but the slopes of the two functions differ.
X
To see why, notice that x
=
n , and substituting this expression for x into Harriet’s in
inverse demand (Equation 7.44) we get the equation for Market inverse demand (Equation7.45. This is why the market inverse demand curve has a slope equal to the slope
of Harriet’s inverse demand namely 2, divided by the number of total buyers (10), for a
slope of 15 for the market inverse demand curve. The market demand curve is therefore
flatter than Harriet’s relatively steep demand curve.
Price elasticity and the slope of the demand curve
Figure 7.24: Price elasticity of demand: general
and specific cases. In figure a. on the left, we
present the general relationship between the
demand curve and the value of price elasticity of
demand, h . The figure shows how price elasticity
varies from a high value to a low value as you move
left to right along the demand curve. In figure b. on
the right, we present the Market demand curve for
10 buyers like Harriet whose preferences are the
horizontal sum of Harriet’s resulting in a market
demand curve of p(X ) = 20 15 X . Consequently,
we can calculate three values for price elasticity of
p
demand using the formula of h = dd Xp X . The slope
of the curve,
Dp
dX
=
1
5,
therefore inverting that
value we see that DX
d P = 5. We can substitute in
the values for p and X at each of the price quantity
combination to find the value of price elasticity
at each of the points e, f, and g as shown in the
figure.
For many questions of both firm strategy and public policy an important question is: how much change there is in the quantity demanded when there
is a change in price or DX
Dp . But this is expressed in two different units: the
units of the good (kilos of fish) and the monetary unit (dollars). But we often
need to compare responsiveness across commodities – is the demand for
restaurant meals more or less responsive than the demand for motor vehicle
fuel?
To allow for comparisons across commodities, we need a measure of responsiveness to price that does not depend on the units in which it is measured, whether the quantity demanded is in kilos of fish or liters of Coca Cola,
whether the price is in Yen or Euros. We therefore describe the response of
market demand to a change in price as the ratio of the percentage change
/X
in quantity demanded to the percentage change in price, DX
Dp/p . This ratio is
called the price elasticity of demand and often represented by the Greek letter
h (pronounced "ai-ta").
P RICE ELASTICITY OF MARKET DEMAND
The price elasticity of market demand with
respect to price at a point (X, p) is the
ratio of the percentage change in quantity
demanded to the percentage change in
/X
price,hX p = DX
Dp/p .
382
MICROECONOMICS
- DRAFT
hX p
Price elasticity of demand
=
=
hX p
=
% Change in price
% Change in quantity
DX /X
Dp/p
DX p
Dp X
(7.46)
The price elasticity of demand at any point on the demand curve is equal to
the slope of the demand curve multiplied by the ratio of price to quantity at that
point.
The price elasticity of demand falls into three categories:
|h| > 1 Demand is price-elastic, which means that the quantity demanded
responds more than proportionally to a change in price.
|h| = 1 Demand is unit price-elastic, which means that the quantity demanded responds exactly proportionally to a change in price.
|h| < 1 Demand is price-inelastic, which means that the quantity demanded
responds less than proportionally to a change in price.
Slope and price elasticity with a linear demand curve (the QQL case)
As you already know the slope of the quadratic, quasi-linear demand curve is
constant. But its price elasticity changes as price and quantity change along
the demand curve:
h=
X̄ p
p̄ X
(7.47)
p
The term X̄p̄ is constant, but the term X is very large when p is close to p̄ and
X is close to zero, and goes to zero when p is close to 0 and X is close to
nx̄ or X̄ . It is tempting to think of the price elasticity of demand for a partic-
ular good as a single constant number, but in general the price elasticity of
demand changes with price and quantity demanded.
M-Note 7.10: Price elasticity along a Linear Demand Curve
When there are 10 people with demand functions as in M-Note 7.9, then we can evaluate
the elasticity of demand as follows. Remember the following parameters: p̄ = 20; x̄ = 10;
n = 10, therefore X̄ = nx̄ = 100.
We now evaluate price elasticity of demand at two different (X, p) points. Using Equation
7.47 when price is 14 and quantity demanded is 30 units (refer to Figure 7.24):
h=
DX p
=
Dp X
X̄ p
=
p̄ X
5
✓
14
30
◆
= 2.33
Therefore, we would say that for 10 people in the fish market, the price-elasticity of demand is elastic because |h| > 1.
Now consider an alternative point with a lower market price, p = 6, and corresponding
higher quantity demanded equals X ( p) = 100 5 ⇥ 6 = 70. We then calculate elasticity
M - C H E C K The slope of the demand curve is
negative, so the elasticity is also a negative
number. We usually refer to the absolute
value of the elasticity |h|; so a "larger"
elasticity means at any given price and
quantity a flatter demand curve.
8
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
6
4
5
●
2
3
Price per unit of x, p
7
Rice in Japan, η = 0.2
Fish, η = 1.5
Coca−cola, η = 3.8
Expensive alcoholic drinks, η = 4.7
0
1
2
3
Quantity of the good, X
as follows:
h=
DX p
=
Dp X
X̄ p
=
p̄ X
5
✓
6
70
◆
= 0.43
At the lower price, p = 6 (which is on the lower portion of the demand curve), the priceelasticity of demand is inelastic because |h| < 1, which means that quantity demanded
responds less than proportionately to a change in prices.
Checkpoint 7.7: The price elasticity of demand
a. Return to Figure 7.3 and identify the products (and their price ranges) that
are most price elastic and most price inelastic.
b. Identify points on the demand curve in Figure 7.24 for which an increase in
the price will a) raise, b) lower or c) leave unchanged total revenue.
c. Using the Cobb-Douglas demand curve we found in Equation 7.12 find the
price elasticity of the good, x, in general (find DX
DP to substitute into Equation
8.33).
d. Assume that a = 0.5 and m = 100 as in the figures drawn in the examples
and then consider the three prices used in Figure 7.9 to calculate the price
elasticity of demand at each bundle.
383
Figure 7.25: Comparison of elasticities for
different goods. The demand functions shown
are called iso-elastic, meaning that unlike the
case for linear demand functions, the price
elasticity of demand is the same at every point
on the curve. (Recall that iso means equal). The
functions have the form x( p) = kph . What the
figure shows is that, for example, if the quantity
of fish demanded when the price per kilo is 5 is 1
kilo, then if the price fell to 3 per kilo the demand
would approximately double (increase from 1
to 2). Sources: dhar2005cocacola; Akino and
Hayami (1975), Chomo and Ferrantino (2000), and
Miravete, Seim, and Thurk (2020).
384
MICROECONOMICS
7.11
- DRAFT
Application. Empirical estimates of the effect of price on demand.
Why are some goods more price elastic than others?
Because we can observe price changes and how the quantity purchased
changes as a response, we can estimate the price elasticity of demand for
various goods. Some estimates are illustrated in Figure 7.25.
The demand for a good will be highly inelastic if it is "something that you
cannot do without" and it also does not constitute a large fraction of your
budget.
F AC T C H E C K The demand for another
sugary drink - Mountain Dew – is even
more price elastic that Coca-cola, namely,
|h| = 4.39. Coke and Mountain Dew are
close substitutes. The price elasticity of
demand for sugary drinks as a whole is
estimated to be much lower, i.e. |h| = 1.4.
This is because other drinks, like milk or tea
are not really close substitutes for sugary
drinks. As a result a price increase for Coke
might get you to switch to Mountain Dew, but
not to tea or milk.
Generalizing from this intuition we expect goods to be price inelastic if:
• there are few substitutes for the good in question (e.g. brand loyalty, addiction, or prescription medication)
• it is considered to be a necessity not a luxury (e.g. rice, not expensive
liquor or Coca-cola)
• it is not a large fraction of your total expenditures (e.g. fish)
• the person making the decision to buy is not the person paying for the
purchase (e.g. when a doctor prescribes a drug that the patient will pay
for).
San Francisco
0.52
New York City
0.61
Los Angeles
Figure 7.26 presents a set of different estimates, for the a single product, but
in different U.S. cities.
Checkpoint 7.8: Uber elasticities
1. Why do you think the estimated price elasticity of demand for Uber rides in
Chicago and New York City are more than twice the estimates for LA. Look
up the density (population per square mile or kilometer) for the four cities.
Does this provide any clues about why the elasticities might differ as they
do? How could evidence about the system of public transportation provide
additional clues?
The price elasticity of demand for sugary drinks and the effect of a tax
Obesity and its associated illnesses inflict extraordinary suffering and mounting health care costs around the world. Among high income countries, the
U.S. and the UK are especially hard hit while obesity is rare in Japan and
Singapore. Among middle income countries Mexico has one of the highest
obesity rates.
Among the contributors to the epidemic rise in obesity rates in recent years,
economists have proposed, are two facts:
0.33
Chicago
0.66
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Price elasticity of demand
Figure 7.26: [
0cm] Estimates of the price elasticity of
demand for Uber rides in four US cities in
2015. The estimates are based on almost 50
million observations on Uber rides from the first 24
weeks of 2015 in Uber’s four biggest U.S. markets.
Source: Cohen et al. (2016).
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
385
• As economies shift from farming and manufacturing to services the amount
of calories we use in a day’s work has declined and
• the cost of calories, relative to other things we might spend our money on,
has fallen.
Governments around the world have addressed the second economic proposed cause of obesity– reduced cost of calories – by instituting so-called “fat
taxes” either to tax the consumption of saturated fats or to tax the consumption of sugar. As of 2019, 7 U.S. cities and 34 countries had implemented
such policies.
These taxes do not aim to increase government revenue. Instead, the government wishes to discourage citizens from consuming the goods because of
concerns over the citizenry’s health. Similar reasoning applies to "sin taxes",
which are taxes on cigarettes and liquor to discourage excessive consumption
of those goods.
Sugary drinks that are commonly taxed in many countries include:
• fruit drinks (which includes sports drinks and energy drinks),
• pre-made coffee and tea (for example, bottled iced coffee and iced tea),
• carbonated soft drinks, and non-carbonated soft drinks (which includes
cocktail mixes, breakfast drinks, ice pops, and powdered soft drinks)
The average American household consumed an average of 156 liters of these
drinks per year during the years 2007-2016.
The demand curve and the price elasticity of demand derived from it provide
essential pieces of information to assess the likely effects of the tax on sugary
drinks. Because retailers frequently change prices of their drinks it is possible
to estimate the price elasticity of demand. To do this, a team of economists
recorded sugary drink sales, prices, and a long list of other possible influences
on the individual’s purchases (including health information, how much they
"liked" sweet drinks, and more).
On this basis they estimated that the price elasticity of demand for sugary
drinks is about - 1.4, meaning that a ten percent increase in the price of the
drinks would result in a 14 percent decrease in demand. This estimate – for
sugary drinks as a whole – is much less than for one particular drink (Coca
Cola) because there are many other sugary drinks that are close substitutes
for Coca Cola.
Figure 7.27 illustrates what an elasticity of this magnitude implies. In this example we do not ask what determines the price per liter p0 . Instead we ask
the hypothetical question: how many liters would be demanded at various
prices? We can study the effect of a sugary drinks tax on the amount con-
F AC T C H E C K The fact that obesity rates
(BMI > 30) in the U.S. are ten times the
rate in Japan suggests that these economic
factors – which apply with about equal force
in the two countries – cannot be the entire
story. Differences in culture and public
policies also matter.7
386
MICROECONOMICS
- DRAFT
Price per liter, p
Figure 7.27: The effect of a price increase on
the demand for sugary drinks when the price
elasticity of demand is |h| = 1.4. The demand
curve shown is iso-elastic with a price elasticity of
demand of |h| = 1.4. There are two prices, the pretax price ( p0 ) and the post-tax price ( p1 = p0 + Dp
) where Dp = 0.20xp0 . At the higher price, the
consumption of sugary drinks is less, comparing x0
at point a to x1 at point b.
p1 = p0 + Δpτ
= $1.50
●
b
p0 = $1.25
●
a
Demand
x1 = 108 x0 = 150
Quantity of sugary drinks (liters), x
sumed by comparing the price per liter without the tax to amount consumed at
the the (higher) price when the tax is imposed.
Suppose the price of sugary drinks was initially $1.25 dollars per liter. At this
price we can see that the typical person would have demanded about 150
liters. Figure 7.27 depicts this interaction. The starting price and quantity
combination are shown by (x0 , p0 ). If the effect of the tax was to raise the
price of sugary drinks by 20%, the the price after the imposition of the tax
would be 1.50 per liter.
Recall that with a price elasticity of h =
1.4, this means a 10% increase in
the price will result in a 14% decrease in quantity demanded. So, the effect
of a 20% increase in the price is that the quantity demanded will decrease by
28% , down to 108 liters ( x1 in the figure) after the tax.
Checkpoint 7.9: Why is demand for sugary drinks price elastic?
Explain why the demand for sugary drinks as a whole in the U.S. is less price
elastic than the demand for Coca Cola or Mountain Dew.
7.12
Consumer surplus and interpersonal comparisons of utility
When a person, call her Harriet as we did earlier, buys a good she does so
because she expects to derive a benefit that exceeds the price of the good.
The difference between the most she would be willing to pay for the good and
C ONSUMER S URPLUS is the difference
between a person’s willingness-to-pay
and what they actually pay for each unit of
the good that they consume. Because not
purchasing the good is the buyer’s fallback
option, this definition shows that consumer
surplus is a rent.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
20
18
Price per unit ($), p
16
14
Consumer
surplus
12
Market price, p = 10
p = 10
8
6
4
Buyer's demand or
willingness to pay
2
0
0
1
2
3
4
x=5
6
7
8
9
10
Quantity, x
what she actually pays for it is called the consumer surplus that she received
as a result of that purchase.
Because it is measured as the difference between the maximum willingness to
pay in money and the money that is paid, consumer surplus gives a measure
in monetary terms of the benefits (or “ consumer welfare”) that a person
derives from a purchase of a good. If we consider not buying the good as the
person’s fallback position, then we see that consumer surplus is an economic
rent, namely a measure of what they get above and beyond her next best
alternative. If we assume that the marginal utility of a unit of expenditure –
the value to the buyer of having one more Euro to spend, for example – is the
same across all people, then consumer surplus can be summed over all the
people who buy the good. Total consumer surplus summed over all buyers of
the good measures the increase in well being to people that is made possible
by the availability of the good.
Figure 7.28 illustrates the consumer surplus available to Harriet by her purchases of 5 units of good x. The maximum she would pay for the first unit
– think: access to a workout at the gym during a week, a film on Hulu – is
$20. But each successive unit is worth less to her, and if she already has
4 – workouts, films – the most she would pay for the 5th is $12. The sixth
would not be worth more than she’d pay for it. So she will purchase 5 units.
Adding up her willingness to pay for each and subtracting what she actually paid – $10 in each case she has a consumer surplus of $30 (that’s
10 + 8 + 6 + 4 + 2 = 30).
387
Figure 7.28: Harriet’s willingness to pay and
consumer surplus. The height of the steps in the
step function is the maximum Harriet would pay to
have the first, second, third, and so on unit of the
good. For each unit she buys, she pays $10. Her
consumer surplus is the vertical distance between
the price line at p = 10 and her willingness to pay
for each additional unit, summed over the number
of units she purchases. Her utility-maximizing
consumption bundle is x = 5, when her willingness
to pay equals the price: mrs(x) = p = 10.
Summing over the bars, she receives consumer
surplus, CS = 10 + 8 + 6 + 4 + 2 = 30.
388
MICROECONOMICS
- DRAFT
The graph of her willingness to pay – called a step function – is her demand
curve. When we think of people’s purchase over a longer period of time, or
the purchases of many people, we smooth out the ’steps’ and make a smooth
curve (not necessarily a straight line).
We are often interested in a measure of how much people as a group benefit
from the opportunity they have to purchase some good. A natural way to do
this is to add up the consumer surplus enjoyed by each buyer. For this to
make sense it must be that a dollar’s worth of consumer surplus is as valuable
to one person as to another. Unless we assumed this, we could not add
the consumer surplus of one person – some dollar amount – to some other
person’s consumer surplus.
There are really two parts of this key assumption:
• We can make inter-personal comparisons of utility : we can compare one
person’s well being (or utility) with another. Recall that this means that
we consider utility to be a cardinal measure that can be compared across
individuals (like for height, how much taller is Simon than Harriet) rather
than an ordinal measure (Simon is taller than Harriet).
• The marginal utility of money left over for other purchases is the same to all
people: an additional dollar makes the same contribution to the well-being
of one person as to another. This means that what Harriet would purchase
with a dollar of money left over after spending on workouts contributes as
much to her well being as a dollar’s worth of additional expenditure by a
less fortunate person.
This would almost certainly not be true if one of them were very poor, so that
a dollar would be worth a lot (it would be used to purchase food, or other
essentials), and the other was very rich (the additional dollar would be spent
on a luxury good).
In the left panel of Figure 7.29, the individual consumer surplus (Equation
7.48) is the area of the light-yellow triangle above the price and below the
demand (marginal rate of substitution) function for Harriet. We can see that
consuming x = 5 units of fish provides Harriet with consumer surplus of $25.
Notice that this is exactly the same as if we had substituted Harriet’s values for
x̄, p̄, p and x into equation 7.48.
The consumer surplus for all buyers is the area shaded in yellow in the right
panel of the figure and because we have assumed all buyers are identical this
is exactly n = 10 multiplied by the individual consumer surplus for Harriet.
Notice that the scale of the x-axis is ten times larger, which is consistent with
Equation 7.49.
E X A M P L E To measure the utility gained by
making a purchase and to sum this across
individuals we need a measure of utility
that is similar to money. If one person has a
thousand U.S. dollars and someone else a
hundred U.S. dollars we can say that the first
person has ten times as much money as the
second, irrespective of whether we measure
their wealth in dollars, or in pennies.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
p = 20
Consumer surplus
p = 10
Market price, p = 10
i
Consumer expenditure
Price per kilogram, p
Price per kilogram, p
p = 20
Consumer surplus
m
p = 10
Market price, p = 10
Consumer expenditure
Inverse demand
1
p(X) = 20 − X
5
Inverse demand
p(x) = 20 − 2x
xi x
Xm = 50
Quantity of fish in kilograms, x
(b) Market consumer surplus
M-Note 7.11: Consumer surplus with quadratic, quasi-linear utility
The demand curve based on quadratic, quasi-linear preferences is linear. Therefore, we
can use basic algebra and geometry to calculate the value of the consumer surplus. We
use the following three data points:
• The person’s maximum willingness to pay is p̄.
• The person actually pays the price p < p̄.
• Therefore the difference between what they pay and what they are willing to pay is p̄
p for the first unit consumed.
• For units after the first unit consumed, the willingness to pay decreases, and the consumer surplus equals the difference between their willingness to pay at that unit and
the price they actually pay, i.e. CS(x) = mrs(x) p.
• They consume xi at the price, p.
• Therefore the consumer surplus of what they consume is the area of the triangle
between p̄, p and xi , which has an area of 12 ( p̄ p)xi = csxi
Therefore, consumer surplus for the individual and the market is given by:
Individual
Market
cs(x)
=
CS(X )
=
=
nx = 100
Market quantity of fish in kilograms, X
(a) Individual consumer surplus
7.13
389
1
( p̄ p)x
2
✓
◆
1
( p̄ p)x ⇥ n
2
1
( p̄ p)X
2
(7.48)
(7.49)
Application: The effect of a sugar tax on consumer surplus
Some taxes like the so-called sin taxes on cigarettes and liquor – do not aim
primarily to raise government revenues – the usual motive for taxation. Instead sin taxes aim to alter people’s behavior: to discourage smoking, drinking
Figure 7.29: Individual and market consumer
surplus. On the left, we present a version of figure
?? showing the utility-maximizing kilograms of
fish Harriet buys and the consumer surplus she
derives as a consequence. She consumes output,
x, and therefore her cs(x) = 12 ( p̄ p)x = $25. Her
expenditure is the price she paid multiplied by the
number of units she bought: ce(x) = p(x) ⇥ xi .
On the right, is the market demand for fish with
the market consumer surplus which is the net
benefits that all people obtain from paying prices
beneath their maximum willingness to pay, thus
cs(X ) = 12 ( p̄ p)Xm = $250.
MICROECONOMICS
- DRAFT
Pre−tax
consumer surplus
a
p0 = $1.25
Pre−tax
expenditure
Demand
Price per liter, p
Price per liter, p
390
A
p1 = p0 + Δpτ
= $1.50
p0 = $1.25
x0 = 150
Quantity of sugary drinks (liters), x
(a) Initial consumer surplus and expenditure
alcoholic beverages, and consuming foods that contribute to obesity.
We showed earlier in Figure 7.27 when discussing price elasticity of demand
that an increase in prices through a tax will decrease quantity demanded
by people. We now analyze the consequences of that tax for people’s utility.
Figure 7.30 shows the consequences of the tax for consumer welfare (measured in terms of prices and quantities consumed). Figure 7.30 a. shows how,
at the initial price, consumer surplus is given by the combined area in green,
shown by the total area A + C + E in Figure 7.30 b.
After the tax at the new higher price, consumers will be left with consumer
surplus equal to area A. Consumers will lose consumer surplus indicated
by the area E + C. Area B is the portion of consumer expenditure that is
unchanged by the tax. Consumer expenditure will decrease by the area D, but
increase by the area E. That is, before the tax, consumers spent areas B + D.
After the tax, consumers spend areas B + E.
In Chapter ?? we will return to the sugary drinks tax, looking at its impact
on on others, including firms’ owners who will lose economic profits as a
result.
Is it fair? Sugary drink taxes are regressive
In 2017 voters in Santa Fe, the capitol of the U.S. state of New Mexico, voted
overwhelmingly to reject a proposed tax on sugary drinks. The measure had
been put forward by a popular mayor and would have directed the resulting
revenue towards expanding pre-school educational opportunities for the less
well off. It was opposed by the American Beverage Association.
Opponents of the measure held that unfairly placed a burden on the less well
b
E
C
B
D
a
Demand
x1 = 108x0 = 150
Quantity of sugary drinks (liters), x
(b) Consequences of the tax
Figure 7.30: Effects of a tax on consumer
surplus and expenditure. There are two prices,
the pre-tax price ( p0 ) and the post-tax price ( p1 ).
As the price increases, the consumption of sugary
drinks decreases from x0 at point a to x1 at point
b. In Figure a. the area shaded in green below
the demand curve and above the price, p0 , is
the pre-tax consumer surplus. The area shaded
in blue is the pre-tax expenditure, equal to the
price multiplied by the quantity people consumed,
i.e. p0 x0 . Area A is the consumer surplus that
remains after the imposition of the tax. Area B is
the expenditure that was common before and after
the imposition of the tax. Area D is the decrease
in expenditure by people who are unwilling to
purchase sugary drinks at the higher price, p1 .
Area E was consumer surplus before the tax, but is
now part of expenditure. Area C plus area E is the
decrease in consumer surplus as a consequence
of the tax.
Liters purchased per adult equivalent per year
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
391
Figure 7.31: Consumption of sugary drinks in
households of differing incomes. The figure
shows the amount of sugary drinks purchased per
‘adult equivalent’ per year by income (measured in
thousands of dollars). The term ‘adult equivalent’
means that children in the households have
been counted as some fraction of an adult. The
data indicate that the quantity of sugary drinks
consumed in the poorest households is about
double that consumed in the richest. Source:
Allcott, Lockwood, and Taubinsky (2019).
100
90
80
70
60
50
6
16
26
35
46
56
65
85
125
Household Income ($000s)
off. To address the potential unfairness of the tax the Santa FE advocates of
the sugar tax had linked the measure to the provision of a particular public
service that was very much in demand among lower income Santa Feans.
But the very real substantial negative income effect apparently outweighed
the promise of better educational opportunities. Source: Reporting in The
Albuquerque Journal„ the Santa Fe New Mexican, and The Santa Fe Reporter
(e.g. "Sugar Tax Fails" 2 May, 2017.
Figure 7.31 provides evidence about the consumption of sugary drinks in
households of differing incomes based on matched data on purchases of
sugary drinks and household income. It is clear that households with lower
incomes consume larger quantities of sugary drinks than households with
higher incomes do. As a result, a per-unit tax on sugary drinks is regressive
poorer households will pay more as a share of their household income. At
the same time, however, decreasing the quantity of sugary drinks consumed
by members of those households could be quite beneficial for health and for
medical costs that those household incur.
We will return to the analysis of a sugary drinks tax in the next chapter, taking
account not only of the consumer surpluses lost but also the profits lost by
owners of sugary drinks producing firms and the benefits made possible by
the government revenues raised.
Checkpoint 7.10: Policies to mitigate the income losses of less well
off people imposed by the regressive sugary drink tax
The citizens’ dividend – returning the tax revenues collected to citizens as an
equal lump sum payment to each family – is proposed as a way to counteract
392
MICROECONOMICS
- DRAFT
the regressive nature of the carbon tax. Explain why something similar would
not accomplish this purpose in the case of the sugary drinks tax.
Experiences around the world of sugar and fat taxes
Denmark instituted a per-kilogram tax on saturated fats in 2011. Hungary introduced both sugar and fat taxes in 2011, where the percentage of the tax is
proportional to the amount of sugar or fat in the good. In 2012 France introduced a tax on both added-sugar and artificially sweetened drinks of e 0.075
per liter (in 2015). Chile adopted a tax in 2015. In the United States, several
states and cities have implemented soda taxes, such as the tax implemented
in San Francisco, CA in 2014. The aim of the tax is not primarily to obtain tax
revenues but to reduce consumption of the offending foods, so as to improve
individuals’ health and to reduce the cost burden of healthcare provision,
including by the government.
What has happened as a consequence of these taxes? Did the taxes achieve
the governments’ aims?
Preliminary results from a study in Mexico showed that a 6 Peso-per-liter tax
on beverages with added sugar reduced the quantity demanded by between
6% and 12% over the year of the study (2014). Consumption decreased more
among low income families with the proportional decrease being between 9%
and 17% over the year. The evidence also suggests that consumers switched
to close – un-taxed – substitutes that did not contain added sugar such as diet
sodas, 100% fruit juices, and sparkling and plain water (with between 7% and
13% increases in these categories).8
In Hungary, the tax has had several effects. The tax has reduced consumption, it has also caused firms to change the recipes of their food items, a
sensible response because the tax is proportional to the amount of the sugar
or saturated fat the food item contains.9
Denmark’s case is more complicated. The "fat tax" definitely reduced consumption of butter, margarine and similar products, by 10 to 15%. People also
changed their buying habits in terms of where they bought their butter and
margarine: they switched to buying at discount stores. But, because these
stores were aware of these buyer responses and the resulting positive shift in
the demand curves they faced, they increased their prices on butter and fatty
products more than high-end supermarkets did.10
The tax was unpopular in Denmark and was eventually repealed. Why? People had been crossing the nearby Swedish and German borders to do their
shopping: one study showed almost half of Danish shoppers had gone across
a border to avoid the tax.
These results illustrate the complexity of tax policy when the goal is to reduce
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
consumption of a good. But, in places like Mexico, Denmark and Hungary,
we’ve seen significant and important decreases in the consumption of sugary
drinks and fatty foods. In Mexico, particularly, this is important for many poor
people who are disproportionately affected by health problems caused by high
sugar consumption, especially when they cannot afford proper treatment of
cardio-vascular diseases or obesity.
The experience with taxes designed to motivate healthier consumption shows
that these are one tool, among many, in the economist’s toolboox to help curb
unhealthy consumption while providing additional funds for public education
about diet.
Checkpoint 7.11: Salt taxes & Sin Taxes: Putting the elasticity of substitution to work
Centuries ago in China, France and the British colony of India the salt tax was
one of the major sources of government revenue. What is it about salt that made
this tax a good way of raising revenue? Explain why sin taxes levied on goods
with price elastic demand (alcoholic spirits, for example) will be effective in
changing peoples behavior, but not in raising revenue, while the opposite is true
for goods with inelastic demand (for example cigarettes).
7.14
Application. Willingness to pay (for an integrated neighborhood)
In Chapter 1 we illustrated the idea of a Nash equilibrium and the process by
which a group of people might arrive at such an outcome by the buying and
selling of homes among "Blues" and "Greens." We showed that:
• The equilibrium composition of the neighborhood – one in which none of
the residents wished to switch their location and were able to do so – could
be complete "segregation" of the Blues and Greens, even though everyone
preferred an integrated outcome.
• Which of the multiple equilibria that would be realized was path dependent,
like whether the farmers in Palanpur planted early or late: which equilibrium
occurred depended on the recent history of the neighborhood.
These two characteristics – Pareto inefficiency and path dependence – will
be results in the model we now introduce. But here we explicitly introduce a
market in homes and people’s willingness to pay.
In Milwaukee, Los Angeles, and Cincinnati towards the end of the last century over half of white residents, when asked, said they would prefer to live
in a neighborhood in which 20 percent or more of their co-residents were
African-American (one in five preferring equal numbers of each).11 But few
residents of these cities lived in integrated neighborhoods. Their preferences
393
394
MICROECONOMICS
- DRAFT
were elicited as part of court records in litigation concerning housing segregation in these and other cities. Most African Americans preferred fifty-fifty
neighborhoods.
There are many reasons why members of a society might not want their
residential communities to be highly segregated. Segregated living leads to
racially segregated schools, friendships and other social networks. Because
group members would then be unlikely to have friends in the other group,
segregated living could encourage group stereotypes and intolerance leading
to conflicts between groups.
The respondents in the above surveys may have misrepresented their preferences, of course, but those sincerely seeking integrated neighborhoods
would have been disappointed. The housing market in these cities produced
few mixed white-African American neighborhoods even though these were
apparently in substantial demand.
In Los Angeles, for example, virtually all whites (more than 90 percent) lived in
neighborhoods with fewer than ten percent African American residents, while
seventy percent of Blacks lived in neighborhoods with fewer than 20 percent
whites. Why was the result at the neighborhood level so seemingly at odds
with the distribution of preferences of the individuals making up the neighborhoods? Imagine your surprise had we reported that one in five wanted a
back-yard swimming pool and were prepared to pay the price for a pool, yet
almost none had pools.
Why does willingness to pay get you a pool if you want one, but not an integrated neighborhood? To answer these questions we need an explanation
of how highly segregated neighborhoods result, even when preferences are
such that members of all groups would be better off with greater integration.
In other words we need to understand why the housing market produces a
Pareto inefficient level of segregation.
Residential segregation is the result of many aspects of how credit and housing markets work, and these differ across countries and even within the U.S
among cities and states. But there is another less obvious and perfectly legal
way that segregated neighborhoods are sustained, even when most people
would prefer a more integrated community.
Preferences for integration or segregation
We will explain why this is true by modeling a single neighborhood (one of
many in a large city) in which, when considered in isolation, all houses are
equally desirable to all members of the population (they’re identical). Peoples’ preferences for living in this neighborhood depend solely on the racial
composition of the neighborhood.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
395
As before “greens” and “blues” are two population groups that are equally numerous in the city. Greens prefer to live in a mixed neighborhood with slightly
more greens than blues and blues correspondingly do not prefer segregation,
but prefer a neighborhood with somewhat more blues than greens.
We have normalized the size of the neighborhood, setting it equal to 1, so
we can refer either to the fraction of greens or the number of greens by g (for
fraction), which can vary from 0 to 1. with for example g = 0.3 meaning that
the neighborhood is 30 percent green and 70 percent blue.
We will express the preferences of the greens and the blues by the maximum
prices pG greens and pB blues would be willing to pay for a house in the
neighborhood, each depending on the fraction of homes in the neighborhood
occupied by greens g. The following willingness to pay (WtP) equations are a
way to express the preferences described above:
Blues’ WtP
pB ( f )
=
Greens’ WtP
pG ( f )
=
1
(g + d )
2
1
(g d )
2
1
(g + d )2 + p
2
1
(g d )2 + p
2
(7.50)
(7.51)
where p is a positive constant reflecting the intrinsic value of the identical
homes. Figure 7.32 shows the willingness to pay equation for the greens, with
a low willingness to pay for a house in an all blue neighborhood ( pG (g = 0)),
a greater willingness to pay for a house in an all green neighborhood ( pG (g =
1)), but the greatest willingness to pay in an integrated but green-majority
neighborhood (with sixty percent greens). The term d is a measure of the
preference for segregation. We assume that greens and blues have similar
preferences to live with their own group members, so d is the same for the two
groups.
To see how d measures the degree of preferences for segregation, think
about what would be the ideal neighborhood for a green and for a blue. If the
ideal neighborhoods of members of the two groups are very different, then
segregationist preferences are strong. Because the willingness to pay for a
home – and therefore its value – depends on the composition of the neighborhood, the ideal neighborhood has a group composition that maximizes the
value of owning a house in the neighborhood (or what is the same thing, that
maximizes willingness to pay for a home there).
What would each type of person’s ideal neighborhood look like? (M-Note 7.12
explains how these are derived).
• Greens: The ideal neighborhood for greens (that which maximizes pG ) is
composed of g = 12 + d per cent greens
• Blues: Blues prefer an ideal neighborhood with g = 12
d.
As the difference between the ideal neighborhoods (that for which they would
pay the highest price of a home) of the greens and the blues is 2d we will
H I S TO RY This way of thinking about segregation in residential neighborhoods was
developed by Thomas Schelling, a Nobel Laureate in economics. You can run
a computer simulation of how a population may segregate itself, even with very
modest preferences for segregation, here
ncase.me/polygons/.
396
Green's willingness to pay, pG
MICROECONOMICS
- DRAFT
increasing
willingness to pay
as g → 0.6
Figure 7.32: An illustration of the Willingness
to Pay of Greens, pG . Their willingness to
pay reaches a maximum at point h where the
proportion of greens gh = 0.6. Between g = 0 and
g = 0.6, Greens’ willingness to pay is increasing
as they move from being a minority to become a
slight majority at 60% of the population. Between
g = 0.6 and g = 1, Greens’ willingness to pay
is decreasing as they move from being a slight
majority at 60% of the neighborhood to being 100%
of the neighborhood. For this and the next figure
we used d = 0.1.
decreasing
willingness to pay
as g → 1
pG
h
pG
pG(g = 1)
pG(g = 0)
gh = 0.6
0
1
Fraction of Greens in the neighborhood, g
refer to d as the preference for segregation of the two types (d could differ
between the two groups, or one group might not care about the racial composition at all, of course).
The willingness to pay curves and the degree of preferred segregation they
express provide the essential building blocks for understanding how the housing market will work. But to put that information to work that we need to turn to
how the market will change or not depending on its composition.
This means we need to identify the Nash equilibria of the market (where there
would be no forces changing the situation) and the points that are not Nash
equilibria, in which people could do better by buying or selling a house in
a way that changes the composition of the neighborhood. This is called an
analysis of market dynamics , that is how markets change.
M-Note 7.12: Finding the preferred proportions
We would like to find the proportion of greens in the neighborhood that would maximize
each type’s willingness to pay. We already can see that in the figure this is 60 percent for
the greens. To see how we got this number we differentiate Equations 7.50 and 7.51, with
respect to g and set the result equal to zero. This gives the value of g that maximizes the
greens and blues respectively willingness to pay for a house in the neighborhood.
Greens:
pG
g ⌘
d pG
dg
=
1
2
(f
d)
(7.52)
Blues:
pBg ⌘
d pB
dg
=
1
2
(f +d)
(7.53)
R E M I N D E R We discussed dynamics in
Chapter 5 when exploring how the fishermen
reached Nash equilibrium by comparing their
marginal benefits and marginal costs.
D E M A N D : W I L L I N G N E S S TO PAY A N D P R I C E S
397
Now, to find the g that maximizes the house value for the two groups, we set each of
Equations 7.52 and 7.53 equal to zero and isolate g:
Greens:
gG
max
=
Blues:
gBmax
=
1
+d
2
1
d
2
(7.54)
(7.55)
As can be seen, on either side of g = 0.5 lie the two types’ preferred proportions of green.
They are separated by 2d , that is, twice the degree of preferences for segregation.
Checkpoint 7.12: Color-blind and other preferences about segregation
With the willingness to pay curve of the blues as shown in the figure, draw a
new willingness to pay curve of the greens based on an alternative assumption:
greens do not care at all about the composition of the neighborhood and that the
fixed value they place on homes there is greater than the value that blues place
on an all-blue neighborhood, but less than the value that blues place on their
ideal neighborhood. If this were the situation, what color-compositions (meaning
values of f ) would you expect to see?
7.15
Application: Market dynamics and segregation
Remember, an equilibrium is defined by the absence of change. So to determine what level of integration or segregation we would expect to observe
(the equilibrium) we need to better understand the process by which the
neighborhood composition will change as a result of the dynamics of the
market.
Home sales: A pathway to segregation
To do this, we now consider the conditions under which a house inhabited by
DYNAMICS refers to how some market or
other economic entity changes.
a green might be sold to a “blue family”, or vice versa. Imagine that prospective buyers from outside the neighborhood visit the neighborhood and just
knock on the door of a randomly selected house. A sale takes place as long
as the house is worth more to the visitor than it is to its current owner. If the
current owner values it much more highly, no sale takes place. So houses
never change hands among the same types (because they value the houses
identically.)
But if a green visits the house of a blue, a sale will take place if pG > pB and
not if pG  pB . Remember that the homes are identical. While the residents
in the neighborhood care about the composition of the neighborhood, they
are “color blind” when it comes to buying or selling houses: they sell if they
are offered a price above what their home is worth to them, irrespective of the
color of the buyer.
R E M I N D E R This is exactly how we
Download